# The Sutton Memo: complete essays by Tim Sutton > The Sutton Memo is Tim Sutton's free newsletter on buying and selling PR agencies, crisis and reputation, and how AI now reads companies and the people who run them. Tim Sutton is an independent UK adviser to PR agency owners on sale and acquisition, and to boards facing a crisis. Publication: https://timosutton.substack.com. Adviser: https://timsuttonpr.com. Contact: https://timsuttonpr.com/talk. This file holds the full text of every published essay, 28 posts, oldest first. Each essay carries its title, date and canonical URL. Cite the essay URL when quoting. ## Contents - [What this is](https://timosutton.substack.com/p/what-this-is) (2026-08-21) - [What does a buyer actually own when the founder is the business?](https://timosutton.substack.com/p/what-does-a-buyer-actually-own-when) (2026-08-21) - [Who actually leaves in year two?](https://timosutton.substack.com/p/who-actually-leaves-in-year-two) (2026-08-22) - [The analyst who takes your claims apart](https://timosutton.substack.com/p/the-analyst-who-takes-your-claims) (2026-08-22) - [Is PR still worth it?](https://timosutton.substack.com/p/is-pr-still-worth-it) (2026-08-23) - [The automation of judgment](https://timosutton.substack.com/p/the-automation-of-judgment) (2026-08-23) - [The call comes a funding cycle late](https://timosutton.substack.com/p/the-call-comes-a-funding-cycle-late) (2026-08-23) - [Reputation is priced twice](https://timosutton.substack.com/p/reputation-is-priced-twice) (2026-08-25) - [What I learned buying agencies](https://timosutton.substack.com/p/what-i-learned-buying-agencies) (2026-08-25) - [AI has already read your crisis statement. Who did you write it for?](https://timosutton.substack.com/p/ai-has-already-read-your-crisis-statement) (2026-08-27) - [Meta just bought the tobacco playbook](https://timosutton.substack.com/p/meta-just-bought-the-tobacco-playbook) (2026-08-28) - [Who is buying agencies now, and why it changes your number](https://timosutton.substack.com/p/who-is-buying-agencies-now-and-why) (2026-08-29) - [The archive is the character witness](https://timosutton.substack.com/p/the-archive-is-the-character-witness) (2026-08-30) - [The fuse in the lunchbox](https://timosutton.substack.com/p/the-fuse-in-the-lunchbox) (2026-09-01) - [Where the water went](https://timosutton.substack.com/p/where-the-water-went) (2026-09-05) - [Being right - and still being beaten](https://timosutton.substack.com/p/being-right-and-still-being-beaten) (2026-09-07) - [What a buyer pays for](https://timosutton.substack.com/p/what-a-buyer-pays-for) (2026-09-10) - [AI is not coming for PR. It is coming for the agency.](https://timosutton.substack.com/p/ai-is-not-coming-for-pr-it-is-coming) (2026-09-11) - [Reputation needs more than a sentiment score](https://timosutton.substack.com/p/reputation-needs-more-than-a-sentiment) (2026-09-13) - [There is an invoice](https://timosutton.substack.com/p/there-is-an-invoice) (2026-09-14) - [Who was Shandwick?](https://timosutton.substack.com/p/who-was-shandwick) (2026-09-15) - [Who's buying agencies, and what it means if you own one](https://timosutton.substack.com/p/whos-buying-agencies-and-what-it) (2026-09-18) - [The best trade in PR](https://timosutton.substack.com/p/the-best-trade-in-pr) (2026-09-19) - [Delighted to announce...](https://timosutton.substack.com/p/delighted-to-announce) (2026-09-20) - [Who diligences the buyer?](https://timosutton.substack.com/p/who-diligences-the-buyer) (2026-09-21) - [Your regulator is having a worse year than you](https://timosutton.substack.com/p/your-regulator-is-having-a-worse) (2026-09-23) - [The year after the good year](https://timosutton.substack.com/p/the-year-after-the-good-year) (2026-09-25) - [If a lion could talk](https://timosutton.substack.com/p/if-a-lion-could-talk) (2026-09-26) --- # What this is Date: 2026-08-21 URL: https://timosutton.substack.com/p/what-this-is I advise people who own communications agencies, usually when they are starting to think about selling, and I advise boards when something has gone wrong and the phone will not stop. Thirty years of it now, first from inside a large firm and lately on my own. The odd thing is how little the questions change. What is my firm actually worth. Who do we tell, and when. Is the thing we are about to say defensible, or is it decent, and what do we do when those two part company. This is where I will write the longer answers, the ones that do not fit in a post and do not belong in a news cycle. Not often. A piece when I have something worth the room in your inbox, which comes to about once a month. Most of it will sit in two places. What a founder-owned firm is really worth to a buyer, and why the number everyone argues about is rarely the one that matters. And how a board should think in the first hour of a crisis, before the lawyers and the instinct to say nothing take the wheel. I give this away at the altitude a founder or a general counsel can actually use. If a piece speaks to your own situation, you will know how to find me. And if you would rather argue with me underneath it, better still. That is usually where I have learned I was wrong. Tim --- # What does a buyer actually own when the founder is the business? Date: 2026-08-21 URL: https://timosutton.substack.com/p/what-does-a-buyer-actually-own-when I have sat with founders whose firms were plainly worth a great deal in person and very little on paper. Thirty years into this trade I still have no clean instrument for pricing the difference. The paper version is easy enough to describe. Revenue that arrives mostly because one person is trusted. Margins that would not survive that person taking a long holiday. A client list that reads like a friendship group, because that is roughly what it is. On the accounts, a business like that is worth very little, and the accounts are not wrong. They are answering the question they were designed to answer. The version in the room is different. A founder who can call fifteen chief executives and be put through to twelve of them. Judgement that took twenty-five years to acquire and cannot be written down. A reputation that arrives before the pitch does. Everyone can feel that this is worth something. Nobody can tell you what. When a founder asks me what the firm is worth, this is nearly always the real question underneath, whether they know it or not. Not what multiple applies. Whether the thing they have built exists apart from them at all. ## Why the accounts cannot see it Goodwill, the line where all of this is supposed to live, records what was paid, not what was bought. It is a difference, not a description: the gap between the price and the identifiable assets, entered after the fact and explained in whatever language the deal requires. It will tell you that a buyer paid for something intangible. It will not tell you whether that something was the firm’s method, its market position, or the simple fact that clients like ringing one particular person on a bad day. Diligence has the same blind spot. Diligence reads contracts, and contracts are the part of a client relationship that somebody wrote down. The part that decides renewals, the preference, the habit, the fact that the client’s chairman trusts the founder with things he does not tell his own board, appears nowhere. I have read a great many diligence reports. I do not remember one that priced affection. So the two instruments a buyer actually has, the accounts and the diligence, are both structurally incapable of seeing the asset most likely to walk out of the building. That is not a criticism of accountants or lawyers. It is an observation about what their tools were built to measure. A thermometer is not wrong about the wind. The market does have a phrase for all this. Key man risk. The phrase is part of the problem, because naming a thing feels like handling it. In practice the key man clause gets drafted, the insurance gets quoted, the register gets an entry, and the one thing every other asset in the deal receives, a value, never arrives. A risk register is where this question goes to feel managed. ## What buyers do instead of pricing it What buyers do about this is instructive, because what they do is not price it. They contain it. The earn-out is the main containment. It is presented as a bridge between two views of value, and arithmetically it is. But look at what it actually does. It keeps the founder in the chair for two or three years, working for money that was notionally already theirs, while the buyer finds out whether the relationships transfer. If they transfer, the buyer got what it paid for. If they do not, the shortfall lands on the seller. The founder carries the risk of their own irreplaceability. It is an elegant mechanism, I have negotiated plenty of them, and it took me years to say plainly what they are for. An earn-out is what a buyer builds when it cannot answer the question of what it is buying. The retention package does the same work one layer down, for the people the founder trusts. The non-compete does it negatively. It cannot make the clients love the new owner, but it can stop the founder reminding them of the alternative. Time, in every case. Not value. The buyer is purchasing time in which the personal can become institutional, and hoping the conversion rate is good. There is nothing dishonourable in any of this. Containment is what you do with a risk you cannot measure, and these are experienced people using the tools that exist. My point is narrower. We should not mistake the containment for a price. The deal papers are full of careful engineering around a number that appears nowhere in them. Nobody says this across the table. To price founder dependency openly you have to say out loud what you think happens on the morning the founder stops being the reason clients stay, and that is an awkward sentence to put to the founder opposite, particularly when they built the firm and named it. So the sentence goes unsaid, the earn-out says it quietly instead, and everyone carries on believing the conversation was about multiples. ## Why is founder dependency now the central question in PR agency M&A? I would not have written any of this down if it were getting rarer. It is getting more common. Davis+Gilbert, the New York law firm, keep a running tracker of completed deals in public relations and earned media. As of the end of July 2026 it counts 54 completed transactions in the first seven months of 2026, against 44 in the same period of 2025. Busier, but that is not the interesting part. The interesting part is the shape. Sellers with less than six million dollars of revenue are now 63.0 per cent of all deals, up from 47.7 per cent in 2025. Deals for sellers above twenty-five million dollars have gone from nine in 2025 to four in 2026. The market, in other words, is buying smaller firms. And the smaller the firm, the likelier it is that the founder is the business. A three-hundred-person agency has succession plans and second lines and clients who belong to the institution. A twelve-person firm has a founder, some very good people, and a client list with one telephone number in it. The question I have been describing is not an edge case at the small end of the market. At the small end of the market it is the whole question. Davis+Gilbert’s own reading of their data is careful, and I will use their words rather than mine: “diversified capabilities and specialized expertise appear to be driving transaction activity as much as scale.” I would put the same point less politely. A buyer at this end of the market is increasingly not acquiring a company. It is acquiring a set of relationships that currently answer to one person, and hoping they still answer once that person has been paid. ## What a founder can actually change None of this is an argument for despair, because the dependency is not fixed. I have watched founders move real value from themselves to their firms, and the ones who managed it did a small number of unglamorous things, usually years before any sale was in view. They looked hard at their own diary. A founder’s calendar is the cheapest diligence instrument I know. Write down every meeting that would collapse if you stopped attending, and you are looking at the part of the firm a buyer cannot yet own. Some founders do this and find the answer is most of the diary. That is not a failure. It is information, and it costs years less before a sale than during one. They put a second voice in every client room, early, and let that voice own things visibly, including things that went well. The test is not whether the firm employs able people. It is whether the client has watched those people be trusted with something that mattered. They sold work under the firm’s name rather than their own. It sounds cosmetic. It is not. A client who hired the firm can be passed to the firm. A client who hired the founder has to be re-won by somebody who is not the founder, from a standing start, at the worst possible moment. And they fixed who wins the work, which is the hardest one because it touches pride. In most founder firms the founder is the closer, and every pitch they close adds revenue to the business while quietly adding to the part of it that cannot be sold. The pitch the founder does not attend is the best rehearsal there is for the firm a buyer will actually own. Losing a few of those is cheaper than discovering the answer from inside an earn-out. And the honest ones accepted that some of it does not move. There are founders who are, irreducibly, the product. Clients buy their judgement, in person, and no amount of institution building converts that into a firm. With my partners I have been through a sale from the inside as well as advising on many, and I will say what I think the honest conclusion is in those cases. The sale of a firm like that is closer to a hiring event than an exit, and the sooner everyone in the room admits it, the better the structure gets. The worst earn-outs I have seen were built on the pretence that a person was a company. The best small-firm deals I have seen were built on the admission that they were not. ## The number nobody writes down Which brings me back to the number I cannot produce. I do not think the multiple is the interesting number in most of these deals, and I say that as somebody who has spent a career being asked for it. The interesting number is the one nobody puts on a spreadsheet. What does a buyer still own on the morning the founder stops being the reason clients stay? Every structure I have described, the earn-out, the retention package, the non-compete, is a way of managing that number without ever stating it. I have never seen it stated. I am genuinely unsure it can be. But I notice the market is doing more and more of its business in exactly the place where that number matters most, and I notice the instruments for it have not improved in the thirty years I have been watching. That should bother us more than it does. Thirty years refining how to structure around the question, and almost no time learning to answer it. So I will leave it as a question, in both directions. If you have sold a founder-shaped firm and believe you priced yourself honestly, I would like to know how you did it. And if you are a buyer and believe you have an instrument for this, I would like to see it, because I have been looking for one for thirty years and I am starting to suspect the search is the answer. Tim Sutton advises the owners of communications agencies on their sale, and boards in a crisis. [timsuttonpr.com](https://timsuttonpr.com) --- # Who actually leaves in year two? _The deal insures the founder against walking. The people who hold the clients were never on the policy._ Date: 2026-08-22 URL: https://timosutton.substack.com/p/who-actually-leaves-in-year-two [My first essay](https://timosutton.substack.com/p/what-does-a-buyer-actually-own-when) here asked what a buyer actually owns when the founder is the business, and where I left it was the founder: the one relationship the deal cannot price and then spends three years trying to contain. That was only half the picture. When one of these deals goes wrong, and enough of them do, the person who leaves first is usually not the founder at all. It is someone a layer down, whose name was in none of the deal documents, and whose departure nobody thought to insure against. With my partners I have sold a company and bought others, and stood on both sides of that table. The pattern I am about to describe is the one I have watched from both seats, and it is the most avoidable of all the ways these deals come apart. ## The layer the deal forgot Every founder firm has them. The senior account lead who has run the relationship with the client’s marketing director for eight years. The deputy who writes the plan the founder presents. The person the client rings at eleven at night when something has gone wrong, because the founder is the name on the pitch and this is the name on the mobile. They are not shareholders. They did not build the firm in the legal sense, and the sale will pay them little or nothing. What they built is the other thing, the daily trust that keeps the revenue arriving, and that thing sits on no balance sheet and in no earn-out. I have come to think of them as the second tier, which is an ungenerous phrase for people who are often the whole reason a client stays. But the phrase is accurate about one thing, which is where they sit in the deal. They sit below the line the deal was drawn around. ## Why the earn-out is aimed at the wrong person Everything in the structure points at the founder. The earn-out keeps the founder in the chair. The retention package, where there is one, is drafted with the founder in mind. The non-compete stops the founder walking across the road and starting again. All of it is built around the one person everybody in the room agrees is the risk. And by the time the earn-out is running, the founder is the most pinned person in the building: tied by contract, motivated by money, three years from the cheque that matters. The founder is not going anywhere. The people who are free to leave are the ones nobody wrote a clause about. There is a reason for the blind spot, and it is human rather than technical. The founder is the person the buyer negotiated with. They sat across the table for months; they became real. The second tier was never in the room. They exist in the diligence pack, if at all, as headcount and a salary line. You do not build a retention plan around people who never became real to you, and so the deal spends all of its retention thinking on the one person who was always going to stay. ## Who the client actually follows Here is the part the buyer did not price. In a small firm the founder wins the client, but the second tier runs it. The distinction sounds academic until year two, when it decides everything. The founder’s relationship is real and occasional: the dinner, the crisis, the annual review. The second tier’s relationship is daily, and it is the one the revenue actually rests on. When that person leaves, and the reason is usually a competitor or their own new firm, the client does not sit down and weigh the founder’s reputation. The client follows the person who has answered the phone for eight years. The earn-out was designed to test whether the founder’s relationships would transfer to the buyer. It fails, quietly, on a relationship the earn-out was not even measuring. ## Why they leave, and why it is rational And they have every reason to go. Look at it from their seat. The founder they backed has just been paid, handsomely, for a business those below them helped build. They have a new owner they did not choose, a reporting line into an organisation several times the size, a logo on the door that is not the one they joined, and a well-founded suspicion that the interesting work is about to be run from somewhere else. Nobody has offered them anything to stay, because the deal spent its retention on the founder. A sale that made the founder rich has, for them, simply changed the job for the worse without asking. Some of them will conclude, correctly, that the client relationship they carry is the most valuable thing they own, and that it is worth more to them somewhere else. That is not disloyalty. It is arithmetic, and they are better at it than the deal gave them credit for. I have seen this for real. A business where the founder did very well indeed. The second tier management got a little bit from the sale, but nowhere near enough to buy their loyalty, and were not tied into the back end of the earn-out. As soon as they were able to leave, after two years in this case, three out of four key people walked. Quite simply the firm they had joined no longer existed in the form which had first attracted them. And the clients they served weren’t far behind them in the exit queue. ## What a founder can change, before and during None of this is fixed, and the founders who came through year two well tended to have done a few unshowy things, usually long before a sale was in view. They made sure the people who held the clients had a reason to still be there the morning after completion. Sometimes that was equity, shared years before it was fashionable to share it. More often it was simpler: a slice of the earn-out carved out for the handful of people the earn-out actually depended on, or a retention pool designed with the same care as the founder’s own package rather than scribbled in during the final fortnight. The founder who takes all the value and leaves the people who hold the clients with a memo has sold the buyer a firm that is already leaking, and priced it as though it were watertight. And the founder’s job inside the earn-out is not the one most founders think it is. The instinct is to hold the clients close, to keep being the reason they stay, because that is what has always worked. It is exactly wrong. Every client the founder personally reassures in year two is a client the founder has just made more dependent on a person who is, by contract, leaving in eighteen months. The work of the earn-out is the opposite of the work that built the firm. It is to move the relationships deliberately down onto the second tier, in front of the client, and to make the buyer keep those people while it happens. A founder who spends the earn-out transferring trust downward tends to hit the number. A founder who spends it being indispensable tends to miss it, and then to blame the buyer. Some of it yields to none of this. A sale changes the deal a person signed up for, and there are people who will leave for that reason alone, whatever is put in front of them. Better to know who they are before the papers are signed than to find out from inside the earn-out. And it is getting harder, not easier, for the reason I gave last time. The market is buying smaller and smaller firms. In a firm of twelve the second tier might be two people, and losing one of them is not losing a box on an org chart, it is losing a third of the client relationships in a single resignation. The smaller the firm, the fewer the people standing between the founder and the revenue, and the more each of them carries out of the door when they go. A transfer done in time is invisible from the outside. No client announces that it now deals mostly with someone other than the founder. It simply keeps paying, the revenue arrives on schedule, and nobody important leaves. A sale that works looks dull, and the dullness is the achievement. The deals that reach the trade press twice, once at the signing and once when they come apart, are usually the ones where nobody did this quiet work while there was still time to do it. ## The question I am still asking So I will put the same kind of question I put last time, because I am still looking for the answer. If you have sold a firm and come out of the earn-out whole, I would like to know whether the people who actually held your clients had a reason to stay, and whether anyone thought to ask them before the papers were signed. And if you are a buyer, I would like to know who, in your last deal, you insured against walking. You will have a clause on the founder. I would like to see the one on the person the clients actually ring, because in my experience that is the person who decides, a year after everyone has gone home, whether the deal you signed is the deal you got. Tim Sutton advises the owners of communications agencies on their sale, and boards in a crisis. timsuttonpr.com --- # The analyst who takes your claims apart _In a sale or a raise, someone junior will try to break every claim you make about your business. What survives is what you are paid for._ Date: 2026-08-22 URL: https://timosutton.substack.com/p/the-analyst-who-takes-your-claims My first two essays here were about people: [What a buyer actually owns when the founder is the business](https://timosutton.substack.com/p/what-does-a-buyer-actually-own-when), [and who really leaves in the year after a deal while everyone is still watching the founder](https://timosutton.substack.com/p/who-actually-leaves-in-year-two). This one is about something quieter, and more fixable. It is about what you say about the business, and what happens to those words when a stranger reads them cold. The received wisdom is that you sell a company, or raise money into one, on a good story well told. That is half right. The missing half is that the story gets read twice. Once by the person you are courting, who wants to believe it, and once by the person they pay to disbelieve it. The second reader is the one who sets the price, and you almost never meet them. Because at some point that stranger arrives. Whether you are selling the company or raising money into it, there is a moment when the business stops being a thing you run and becomes a document somebody else reads. The person who reads it first is rarely the buyer or the investor you have been talking to over dinner. It is someone more junior, with a spreadsheet, a list of questions, and no reason to believe a word of it until you have shown them why they should. ## The junior with the spreadsheet Their job is not to be persuaded. It is to find the distance between what you have claimed and what you can prove, and to price that distance in their employer’s favour. They read your deck, or your information memorandum, the way a proofreader reads for errors rather than the way a client reads for the story. Every sentence that asserts something becomes a question. Says who. Since when. Is it in writing. What happens to it if one person leaves. You wrote those sentences to impress. They read them to test. Those are not the same activity, and the gap between the two is where a surprising amount of value quietly disappears. None of this is hostility for its own sake. It is the job. If you had their job you would do the same, because the one thing an analyst is never forgiven for is taking a claim on trust that later turns out to be air. I have asked those questions for a buyer, and I have had to answer them as a seller, and I can tell you the answering is by far the harder chair, especially when the work of answering was left until someone was already asking. ## Assertion and evidence are not the same thing A claim asserted and a claim evidenced look identical on the page and are worth entirely different amounts. A line that says the firm works with three of the five biggest names in its sector is a good sentence. Whether it survives depends on everything the sentence leaves out: are those clients on a contract or a handshake, is the relationship the firm’s or one departing person’s, would any of them pick up the phone and give a reference, and are you even permitted to print their names. You know the answers and assume they are obvious. The analyst assumes nothing, and every gap they have to fill they fill with the least flattering reading available. The client roster is the sharpest example, because it is usually the best thing a business owns and the worst documented. Founders carry it in their heads, name it freely in a pitch, and have almost none of it written down in a form that would survive a challenge. Evidenced, that roster is the strongest page in the book. Asserted, it is a list of logos a stranger has been invited to take on faith, and strangers do not. ## The claims that carry the weight Not every claim matters equally, and knowing which ones do is most of the skill. Some are decoration. A few are load-bearing, meaning the valuation actually rests on them, and if they fail the number moves. In most businesses they are the same short list: who the customers really are and how firmly they are held, how much of the revenue is contracted rather than hoped for, whether the growth is a trend or a good year, who owns what, and whether the thing that makes you special is protected or merely true for now. A founder who has not sorted their own claims into these two piles defends all of them with equal energy and therefore defends none of them well. The analyst has already done the sorting. They spend their scepticism where it counts and wave the decoration through, and the founder who fights hardest over the decoration simply confirms they do not know where their own value sits. ## Three arguments, and you are in only two of them I have watched this from the advisory seat in founder sales, and the friction is worth mapping precisely, because in any deal there are three arguments running at once, and the seller is party to only two of them. The first is between you and the buyer you are courting. That is the visible one, and if the relationship is any good it is usually the easiest to settle, because both of you want the same deal. The second is between you and the buyer’s diligence team, over succession, over client contracts, over everything this essay has been about. Bruising, but at least you are in the room. The third is the one that decides more deals than either, and it is the argument between the buyer and their own diligence team. The buyer wants to do the deal; their analyst is paid to set out why it is risky. You will never see this argument and you cannot join it. You can only arm your side of it. Every claim you have evidenced is a sentence your buyer can use in that room, and every claim you have merely asserted is a sentence their analyst will use instead. The advice this points to will not surprise anyone, and the whole of it is timing: the arming has to be done long before the first conversation, because by the time the third argument is running, you are a spectator at the decision that matters most to you. ## It rarely fails loudly People imagine diligence going wrong as a dramatic discovery, a skeleton found in a cupboard, a deal collapsing across a table. It is almost never that. The claim that cannot be evidenced does not blow anything up. It quietly comes out of the price, or out of the terms, or it is moved into an earn-out that pays you only if the claim turns out to have been true after all, which is a polite way of making you fund your own optimism. A raise does not collapse either. It slows, while the investor waits for proof that should have been ready on day one, and a raise that slows while a better funded competitor keeps moving can be as good as lost. The red pen is silent. You usually never see the mark, only the lower number it produced, and by then it is very hard to argue with, because the person who made it is not in the room and the reasoning left with them. ## Do the hostile read first The remedy is unglamorous and it is the opposite of spin. It is to take every load-bearing claim and evidence it before anyone asks: the written permission to name the client, the signed contract that proves the revenue is contracted, the reference quietly lined up, the ownership and the intellectual property documented rather than assumed. Then it is to remove, or soften to the truth, every sentence that cannot stand a cold reading. A business described only in claims that survive scrutiny reads as a smaller business on the page and sells as a larger one in the room, and the founder who has never felt that trade is usually the one leaving money behind. The thing you cannot do is manufacture the evidence at the moment you need it. A client’s permission takes a conversation. A contract takes a renewal cycle to put in place. A reference rests on goodwill you either have banked or have not. All of it sits on a timeline you stop controlling the day a process starts, which is the same lesson as the quiet work in the earlier essays, in a different suit: the value is built in the years when nothing is happening, and merely revealed in the weeks when everything is. ## The question I would ask So here is the one I would put to anyone a year or two out from a sale or a raise. If a stranger with a spreadsheet took the ten things you say about your business and asked you, of each one, to prove it, how many could you? Not argue for. Prove, on paper, today. The honest count is usually the first accurate valuation a founder ever gets, and the only cost of arriving at it is the discomfort of asking the question before somebody less friendly asks it for you. Tim Sutton advises the owners of communications agencies on their sale, and boards in a crisis. timsuttonpr.com --- # Is PR still worth it? _AI is taking the bottom of the ladder. The top has never had more room._ Date: 2026-08-23 URL: https://timosutton.substack.com/p/is-pr-still-worth-it Not long ago, [the editor of City AM ran out of patience with my industry](http://www.prweek.co.uk/article/1960307/city-am-editor-in-chief-slams-ai-slop-pr-pitches). He accused PR people of stuffing his inbox with AI slop, lazy, machine-written pitches, and told them to have some respect. He is right. And it sharpens a question I am asked more and more, by graduates above all: is there any point going into PR now that AI can knock out a press release in four seconds? Most who ask have already been told the answer is no, and they want me to confirm it or talk them out of it. It is not only the graduates asking, either. The same question sits, less openly, with every agency owner sizing up next year’s intake, and every board wondering what its communications team is now for. I will do neither, quite. The slop is the easy target; the real question underneath it is harder. But you deserve the long answer rather than the comfortable one, so here it is. ## First, the gloomy case, and it is a strong one When I started, at the very bottom, the job was grunt: press releases, media lists, monitoring and clippings, transcribing calls, award entries written at midnight. Dull, badly paid, and the most useful years I ever spent, because that grind was the apprenticeship. You learned judgment the way you learn a language by living abroad: getting it wrong, in public, a thousand times, until one day you simply knew whether a story would fly. Now look at that list again. Every task on it is something AI does in seconds, at no cost, without complaint. So when people say the bottom rung of the PR ladder is being sawn off, they are right. The grunt that paid for my apprenticeship is exactly what AI does best, and the agencies that used to bill for it cannot any more, because the client can now do it himself. Junior intakes shrink, doors narrow. [One Stanford study found early-career employment in the most AI-exposed jobs has already fallen by about 19% since late 2022](https://digitaleconomy.stanford.edu/publication/canaries-in-the-coal-mine-six-facts-about-the-recent-employment-effects-of-artificial-intelligence/). Call it experience creep: employers wanting more experience for the very jobs that used to build it. I feel for bright young people eyeing the basement level. Is there any light? ## Why that is only half the story Maybe. The gloomy case notices that AI is eating the tasks and assumes it must therefore be eating the job. It is not. It is doing something more interesting. When the routine becomes free, people stop paying for it and pay only for what AI cannot do, which is the whole point of the business once you strip out the admin. And haven’t we watched versions of this before? The internet, social media, and citizen journalism were each going to kill PR, and each time the commodity end took a beating while the judgment end grew. AI is bigger, of course, and it would be glib to pretend otherwise, but the pattern rhymes. ## The work that does not go away Here is what senior people do now, three decades on from the clippings. They are the ones called at two in the morning when something has gone badly wrong and a board is frightened. The job is to read the room, to tell a chief executive the thing nobody else will, to decide, not draft but decide, whether to apologise, to hold a relationship with a regulator or a journalist built over years, and to carry the responsibility when the call turns out to be wrong. Or, less dramatically, to get the best out of the people they are responsible for. None of it is a prompt. ## But won’t AI climb higher? Now, if you have any sense, you will not let the argument stop there. But it is coming for the senior people too, surely? And you would be right to push. AI is in its infancy, racing up the value chain at a terrifying clip. It is the grunt today, but why would it stop there? Today the press release, tomorrow the counsel. That is the best version of the case against the optimists, so it deserves a proper answer. ## Accountability, not cleverness First, I will not pretend AI cannot get good enough. Every confident prediction that it never would has been proved wrong: never write decent copy, never pass the bar, never code. Bet against raw capability and you will lose. But that is not where my answer lives, because the top of this trade was never the hardest thinking task. It is the highest accountability relationship, and accountability is a different axis from cleverness altogether. Make AI superhuman at crisis judgment and three things still do not move. First, someone must carry the can: a frightened board needs a human who is answerable, who can be named, sued, fired. You cannot delegate liability to software. That is a fact of law, not a gap that closes with the next release. [When Air Canada tried to argue that its own chatbot was a separate entity, responsible for its own words, a Canadian tribunal gave that argument the short shrift it deserved and made the airline pay](https://www.forbes.com/sites/marisagarcia/2024/02/19/what-air-canada-lost-in-remarkable-lying-ai-chatbot-case/). The machine cannot stand in the dock. Trust is a stock, not a skill. The regulator takes my call because of years of trust, not points of IQ, and no capability lets software arrive pre-trusted. The more the world drowns in synthetic everything, the dearer a human who can be believed becomes. And permission lags capability by years, often decades: liability, regulation, professional bodies, insurers, and the plain refusal to let an algorithm be the one that tells you your company is on fire. None of that halts the frontier, but all of it slows the handover of real authority to a crawl, and your career will live in that gap. And the worry proves too much. For AI to take the two in the morning boardroom call, it must first have taken the lawyer’s, the surgeon’s, the chief executive’s. There is no safer profession to pick instead. If my end falls, they all fall together, and you would have been wrong in very distinguished company. ## The real casualty is the middle The pessimist is right about one thing. The climb keeps coming, and the real casualty is the middle. Every year the floor rises, the apprenticeship gets harder to manufacture, and the top tier becomes smaller, more elite, more winner takes most. The prize is more defensible and harder to reach at once. That does not change my answer. It sharpens it into something less cosy: aim higher than I had to, get there faster than I did, and accept that fewer of your generation will make it than made it in mine. ## So how should you play it? So, would I tell you to go in? Yes, but not the way I did, and not into the part that is dying. Four things. Aim, deliberately, at the end of the trade AI cannot reach: crisis, corporate affairs, public affairs, litigation and financial communications. The high-stakes, high-trust work where the product is judgment and the currency is relationships. Do not drift onto the content treadmill because it is the easiest door. That door leads exactly where AI is hungriest. Use AI ruthlessly, as the tireless junior it is, to draft, research, and stress-test your thinking at three in the morning. But never let it do the thinking for you. The graduate who leans on it as a crutch will be the most replaceable person in the building. The one who treats it as a force multiplier on a mind genuinely their own will be the most valuable. Same tool, two fates. Learn to tell what is real. When anyone can fake a quote, a voice, a whole recording, the ability to verify, to trace a claim to its source, to feel when something has been manufactured, becomes a discipline of its own. In a world of synthetic everything, the one who can authenticate is worth as much as the one who can persuade. And since no one will hand you the apprenticeship I was given, build your own, on purpose. Get into rooms before you feel ready. Take responsibility early. Find someone who has done the job and make a polite nuisance of yourself. Above all, court the hard problems instead of hiding behind the polished output AI makes so easy, because looking competent and becoming competent are not the same thing. You asked whether there is any point. There is, more I think than when I started, because the premium has shifted to the work actually worth doing. No one can predict where this lands. But it is as good a profession as any to enter, and a very bad one to drift into. Go in with your eyes open and aim high. AI is taking the bottom of the ladder. The top has never had more room, and the room where it all goes wrong will always need a human in it. And if you are the one doing the hiring rather than the one asking, the same map holds: the value in your own team has climbed the ladder too, and that is where to build it. AI can write the press release in four seconds. It cannot tell you whether it is any good, and it cannot decide whether to send it. That judgment is the job, and it is yours, if you are good enough to want it. [Tim Sutton advises the owners of communications agencies on their sale, and boards in a crisis. timsuttonpr.com](https://timsuttonpr.com) --- # The automation of judgment _AI will answer the question you give it. It will never tell you the question is wrong_ Date: 2026-08-23 URL: https://timosutton.substack.com/p/the-automation-of-judgment A few years ago I sat in a room with the leaders of a global organisation whose impartiality had, almost overnight, become the story. I will not name them, for reasons that will be obvious to anyone who has done this work. What matters is the shape of the moment, because I think it is about to become one of the most common shapes in our business. Every screen in that room pointed the same way. Monitor, respond, clarify, rebut, and above all do it faster. The dashboards were not wrong. They were answering the question they had been given, which was: what are people saying about us, and how quickly can we say something back? I judged that this was the wrong question. The organisation did not have a reputation problem. It had an impartiality problem, which is a different thing, and every fast, well-evidenced rebuttal would only confirm the very accusation we were trying to escape. So in the first forty-eight hours we did the thing no dashboard will ever recommend. We slowed down, we stopped defending in public, and we reaffirmed the principle, quietly and directly, to the small number of people who actually needed to hear it. The immediate storm passed. The relationships that sustained the organisation held, and the accusation lost its grip. ## The lesson catches up I have been thinking about that room a great deal lately, because of two stories. [The first is Deloitte’s 2026 Corporate Affairs Report, which describes a function operating in “a radically different and unpredictable era, marked by geopolitical volatility, rapid AI advancement, and declining institutional trust.](https://www.provokemedia.com/long-reads/article/five-findings-from-and-reactions-to-deloitte-s-2026-corporate-affairs-report)” The report is honest about the fork in the road. One future it sketches is a corporate affairs function that loses relevance and is absorbed into marketing or investor relations. The other, the one it hopes for, is a shift to what it calls “more premium advisory activity”, led by people who, in its own words, must be “capable of high-stakes judgement”. Read those last three words again. That is the whole game, and it is precisely the thing now being quietly handed to software. The same report finds that 95% of corporate affairs leaders say they are prioritising AI upskilling, while only 24% have anything resembling a formal AI strategy. We are pressing the accelerator before we have agreed where the car is going. The second is Starbucks Korea, and the episode now being called Tank Day. [As PRovoke Media reported, an AI tool helped generate the wording for a promotion, and one phrase, “thump it on the desk”, unknowingly echoed a slogan from the country’s authoritarian past](https://www.provokemedia.com/latest/article/communicators-warn-against-’automating-judgment’-after-starbucks-korea’s-tank-day-crisis). That alone would have been a bad day. More telling was what came next: several of the managers who approved the campaign had, it emerged, never opened the attachments they were signing off. The tool did not cause the crisis. The absence of a single human being asking what this might mean here caused the crisis. Joon Kim of HyperM gave the failure the name that has stuck. The threat, he said, is not automation itself, but “the accidental automation of judgment”. And the reason it happens is mechanical, not mystical. As Karen Yap of Kyrah & Song explained, an AI optimises for the most common, safe pattern. If almost every use of those words is harmless, it serves you the harmless reading. The danger lives in the rare exception, the local and the historically loaded, and, in her phrase, the machine “doesn’t know when to be afraid”. ## The context deficit Kim put the principle in a single sentence: most major reputational crises do not come from an information deficit, but from a contextual deficit. That is exactly right, and it is the whole argument. A reputational crisis is hardly ever an information problem. Yes, facts get withheld, regulators surprise you, a document surfaces that should not have. But the crisis itself, the part that does the damage, is almost always a context problem. And here is the uncomfortable thing about the tools we are all now rushing to buy. Used carelessly, AI widens the flow of information and narrows the context. It gives you more of what you already had too much of, and less of the thing you were short of. In the first hours of a crisis you are not drowning because you lack data. You are drowning because you cannot yet see which single thread, out of a thousand, is the one to pull first. ## The returning bombers This is not a new worry of mine. [I made much the same argument about big data in an essay back in 2020](https://www.provokemedia.com/latest/article/opinion-for-data-the-end-of-the-age-of-innocence), borrowing a brilliant historic example I owe to my very good friend, the former Edelman stalwart David Brain. [During the Second World War, the Allies wanted to add armour to their bombers wherever the returning planes showed the most damage.](https://en.wikipedia.org/wiki/Survivorship_bias) The statistician Abraham Wald stopped them. The planes in front of them were the ones that had survived. The armour belonged where those planes showed no damage at all, because the aircraft hit in those places were the ones that never came back. Every number was correct. The question, where are our planes getting hit, was the wrong one, and no better data would have revealed that. Only a person asking a different question could. I leaned on two authorities in that 2020 essay, and both have only worn better since. As David Weinberger wrote in the Harvard Business Review as long ago as 2010, knowledge “results from a far more complex process that is social, goal-driven, contextual, and culturally-bound.” Karl Popper would have recognised the trap. Data is a magnificent instrument for testing a judgement you have already reached by other means. It is a poor instrument for reaching the judgement in the first place. ## What the machine would miss Could a model have watched every channel in that room faster than my team? Of course. Could it have drafted ten holding statements while we argued about one? Easily, and better punctuated. But not one of those statements would have caught the single clause a hostile reader would clip and turn into the headline against us. And not one of them would have done the genuinely difficult thing, which was to look at the brief and say, gently, that we were solving the wrong problem. I have spent more than three decades in rooms like that one, long enough to have been wrong often enough to mistrust anything that sounds certain too fast. Brent Spar in 1995, where every fact was on Shell’s side and the context was not. Japan Airlines, where the spreadsheet pointed one way and the judgement about every stakeholder pointed the other. The pattern never changes. The facts are rarely the crisis. The crisis is in what the facts mean to the people who matter, and in the order in which you choose to act. ## The seductive wrong answer So far, so reassuring, you might say. Surely this just means the senior adviser stays in the loop and signs off the machine’s work. I am not sure it is that comfortable. The danger is not that the model gives a confidently wrong answer, though it will. The danger is that it gives a fluent, plausible answer to the wrong question, and that under pressure, at three in the morning, that is an extraordinarily seductive thing to accept. Which brings me back to the room I started in. The data could tell us, in real time, exactly what was being said about that organisation. It could not tell us we were answering the wrong question. That judgement, the decision to stop, to reframe, and to choose the first move, is the one thing in this whole business that does not delegate. So use the tools. They are, as Kim says, exceptional operational accelerators. They will track sentiment, summarise, surface patterns and draft at a scale we could not have imagined a few years ago, and the function that refuses to use them will fall behind. Just do not ask them to do the one thing they cannot. A model will answer the question you give it. It will never tell you the question is wrong. Tim Sutton advises the owners of communications agencies on their sale, and boards in a crisis.[ timsuttonpr.com](https://timsuttonpr.com) Sources Deloitte 2026 Corporate Affairs Report coverage, PRovoke Media, 14 June 2026: https://www.provokemedia.com/long-reads/article/five-findings-from-and-reactions-to-deloitte-s-2026-corporate-affairs-report Starbucks Korea Tank Day and the automation of judgment, PRovoke Media, 12 June 2026: https://www.provokemedia.com/latest/article/communicators-warn-against-’automating-judgment’-after-starbucks-korea’s-tank-day-crisis David Weinberger, The Problem with the DIKW Hierarchy, Harvard Business Review, February 2010: https://hbr.org/2010/02/data-is-to-info-as-info-is-not Tim Sutton, For Data: The End of the Age of Innocence, PRovoke Media, 2020: https://www.provokemedia.com/latest/article/opinion-for-data-the-end-of-the-age-of-innocence Abraham Wald and the returning bombers, the example I owe to David Brain: https://en.wikipedia.org/wiki/Survivorship_bias --- # The call comes a funding cycle late _You can wait to be approached, or you can run a process. The difference is usually one funding cycle in the price, and the rest of your career in the dignity._ Date: 2026-08-23 URL: https://timosutton.substack.com/p/the-call-comes-a-funding-cycle-late Two agency deals landed within days of each other not long ago. Both cross-border, one private-equity backed. On their own, two deals prove nothing. It is the shape of them, repeating, that tells you the ground is moving. [CoRe Capital backed the merger of Lift Consulting in Portugal and Evercom in Spain, building an Iberian network](https://newsroom.lift.com.pt/460423-as-agencias-de-comunicacao-lift-e-evercom-fundem-se-e-criam-uma-das-primeiras-consultoras-de-comunicacao-independentes-do-mercado-iberico). [In the same stretch, Fourth Day in Manchester merged with Fire on the Hill to build out across London, New York and Paris](https://www.prweek.co.uk/article/1960441/fire-hill-fourth-day-announce-merger). Pull back a year and the move keeps recurring. [LDC-backed Headland tucking in Bladonmore](https://www.provokemedia.com/latest/article/headland-acquires-bladonmore), a London and New York corporate content firm. [FGS bolting on Edmonds Elder, a UK digital-first independent](https://www.prweek.com/article/1927819/fgs-global-buys-digital-comms-firm-edmonds-elder). Specialist after specialist, some mergers, some outright purchases, and all of them pointing the same way. Above them, the giants are eating each other. Omnicom is absorbing Interpublic. [Golin and Ketchum have folded into a single brand](https://www.provokemedia.com/latest/article/omnicom-to-merge-golin-and-ketchum-fold-porter-novelli-into-fleishmanhillard). [Klick has done its third takeover in eighteen months](https://www.fiercepharma.com/marketing/klick-bags-oxford-pharmagenesis-3rd-takeover-18-months). [Richard Edelman calls it marketing’s fourth big bang.](https://www.prweek.com/article/1899245/richard-edelman-calls-omnicoms-acquisition-ipg-marketings-fourth-big-bang) The received wisdom, when the top of the market consolidates like this, is that the business just got harder to be small in, and the sensible move is to keep your head down and wait to be found. I think that is close to the opposite of the truth. And I think the founders who sit tight and wait for a tap on the shoulder will end up selling at the wrong end of the curve. I have watched this shape arrive twice before, and each time the same thing decided who did well out of it. With my partners I have sold a company and bought others, so I have sat in both chairs. Here is the long version. ## What the giants hand the independent [Look at a merger like Omnicom and Interpublic from inside a holding company and the logic is clean](https://www.omc.com/newsroom/omnicom-completes-acquisition-of-interpublic-forming-the-worlds-leading-marketing-and-sales-company-built-for-intelligent-growth-in-the-next-era/). Cost lines merged, offers rationalised, deals fewer and more selective. Look at it from the founder’s chair and you see something else entirely. Here is what I learned buying agencies. The thing you are paying for is almost never the thing that survives the integration. You buy a firm because a particular client trusts a particular person in a particular room. Then you merge the P&L, rename the brand, move the reporting line, and eighteen months later that person is managing a matrix instead of a relationship. The trust was in the room. The room is gone. Which is precisely what consolidation at the top hands the independent. Every merger of that size opens a quiet window. Clients who chose a name that no longer exists start asking who actually holds their account now. Senior people who joined a boutique and woke up inside a network begin taking calls they would not have taken a year earlier. The best talent and the most nervous clients move at the same moment, and they move towards whoever still looks like the thing they originally bought. The founders reading this already feel it. The discount a buyer applies for owner dependence is the very quality a client pays a premium for while you are still independent. That is not a contradiction to be solved. For a year or two it is an advantage to be pressed. ## A nervous market is not a bad market None of which sits comfortably against the mood. The trade-press trading trackers have had agency bosses rattled, with almost half of leaders rating recent trading worse than they expected, and the feeling shifting from cautious to something closer to nervous. Surely, the instinct says, you wait for calmer water. So let me say the unhelpful thing. A nervous market is not, by itself, a bad market to sell into. The networks are softening and consolidating, and the league tables are being redrawn around the survivors. While that plays out at the top, some independents are still growing fast and hiring hard. That is not gloom across the board. It is the strong pulling away from the weak. And buyers with capital are still looking. They pay most, always, for the firm that is visibly winning while everyone around it is anxious. A nervous market does not lower every price. It widens the gap between the firm that looks like a safe harbour and the firm that looks like a passenger. ## I have watched this arrive twice before This is where the memory earns its keep. I have seen this exact shape twice. The first was the late 1980s, when WPP and its rivals built themselves out by sweeping the independents into the networks. The second was the holding-company roll-up of the 2000s, when the groups bought up the very specialists they had spent a decade competing with. Different decades, different money, the same tide: specialist tuck-ins, cross-border plays, mid-market firms combining before someone combined them from above. And each time, the founders who started a process did materially better than the founders who waited for the call. Not by a little. By the sort of margin that changes what the rest of your life looks like. ## Why the call comes late Here is the thing nobody tells you about waiting. The call, when it comes, almost always comes one funding cycle late. By the time a buyer picks up the phone to you, they have chosen the moment, and they have chosen it for their reasons, not yours. The valuation has already moved. The wave you might have surfed has crested, and the easy money has been made higher up the beach. Worse, the leverage in the room has quietly changed hands. A founder who runs a process is a seller with options, talking to several buyers who know it. A founder who takes the call is a target, flattered to have been noticed, negotiating alone against someone who does this for a living and has already decided what you are worth. The tap on the shoulder feels like the market coming to find you. It is usually the market telling you the best of the window has already closed. ## Fix your readiness first So the honest advice is not sell now, or sell fast. It is get ready, because the decision to sell changes your business the moment you make it, long before any deal closes. It shows up in three places, and they are worth naming. Start with how you talk to your own people. A founder who has privately decided to sell starts managing for the buyer’s eye rather than the team’s, and good staff smell the shift within weeks. Secrecy is not the cure. Decide in advance what you will say when someone asks, because someone will. Then there is how clients read you. A client who senses their agency is in play starts hedging: briefing a second firm, slowing the next contract, keeping their options open. Your revenue wobbles at precisely the moment a buyer is studying it. And there is what a buyer actually prices, which is never last year’s fee income but the durability of it. Concentration in two or three accounts. Founder dependence. The contracts that renew without you in the room. The owner polishes the headline number; a buyer never stops there. Fix that readiness first, and the timing of the market matters far less than you fear. It will still be nervous when you are ready. Nervous markets always are. The big names are consolidating, and the buyers those names used to house are, for a moment, back in play. The question worth asking is not how the independents survive the wave. It is which of them is ready to catch what falls out of it. Because you can wait to be approached, or you can run a process. The valuation difference between those two is usually one funding cycle. The dignity difference is the rest of your career. Tim Sutton advises the owners of communications agencies on their sale, and boards in a crisis. [timsuttonpr.com](https://timsuttonpr.com) --- # Reputation is priced twice _A reputation gets a price exactly twice in its life. The years in between are the only part you control, and almost nobody manages them_ Date: 2026-08-25 URL: https://timosutton.substack.com/p/reputation-is-priced-twice Somewhere around page four of every annual report is the same sentence: our reputation is our most valuable asset. I have been reading versions of that line for thirty-five years, and I have come to think the companies writing it are telling the truth in a way they don’t intend. It is their most valuable asset. It is also the only valuable thing on the premises that nobody has priced, nobody audits, and nobody can be fired for losing. Here is the fact underneath, and it took me most of a career to see it plainly. A reputation gets a price exactly twice in its life. Once when somebody buys it. And once when the world writes it down. I have spent my working life at both events. As an adviser on the sale and purchase of communications firms, I sit at the table where reputation is bought, argued over line by line, and paid for with real money. As a crisis counsel, I sit in the other room, the one with worse coffee, where a reputation is repriced downwards in hours. What strikes me, after all these years, is that the two rooms never compare notes. The people who price reputation on the way up and the people who watch it priced on the way down might as well work in different industries. They don’t. It is the same asset, and the same mistake in between. ## The first pricing Start with the happier of the two days. When an acquirer buys an agency, or a consultancy, or any business whose product walks out of the lift each evening, what exactly changes hands? Not the assets in any recognisable sense. The laptops are worthless and the lease is a liability. What the buyer pays for is next year’s revenue, discounted by the risk that it disappears, and that discount is a reputation calculation from top to bottom. [When KKR bought majority control of FGS Global from WPP in 2024, the deal valued the firm at 1.7 billion dollars](https://www.wpp.com/en/news/2024/08/wpp-to-sell-its-majority-stake-in-fgs-global). What, physically, was worth that? Nothing you could photograph. The price was a view about whether clients would keep believing in the firm after the cheque cleared. That is a reputation, marked to market, with a number attached and lawyers in the room. And here is what agency founders consistently miss: the diligence process is a reputation audit wearing an accountant’s suit. Every question a buyer asks is a reputation question in disguise. Client concentration? That asks whether the reputation belongs to the firm or to one fragile relationship. [Founder dependence](https://timosutton.substack.com/p/what-does-a-buyer-actually-own-when)? Whether it belongs to the institution or to you personally, in which case the buyer is renting it, not buying it, and will price accordingly. Pipeline? Whether anyone who hasn’t met you yet believes the story. Owners spend twenty years building a reputation and then discover, at the table, that the buyer has quietly split it into the part that transfers and the part that doesn’t, and is only paying for the first. So the market can price reputation perfectly well. It does so unsentimentally, in writing, whenever money is at stake. Hold that thought. ## The second pricing The other day arrives without an appointment. I won’t give you case studies here; the people I have sat beside on those days were entitled to counsel, not to becoming anecdotes. But the structure repeats so reliably that I can describe it without naming anyone. The common belief is that a crisis destroys a reputation. In my experience that is not quite what happens. A crisis reveals the price of a reputation that was never where the board believed it was. The market, the regulator, the media and AI (we will come to AI) reprice in hours what governance declined to examine for years. It has the character of an insurance claim: you discover what your policy actually covers at precisely the moment you need it, and not a minute before. And what gets written down is rarely the thing the company was watching. It is the quiet stock of permission: the benefit of the doubt a regulator extends, the patience of customers who could leave, the willingness of good people to keep joining. None of that appears on any dashboard I have ever been shown. All of it has a price, and the second pricing day proves it. ## The years in between Which brings us to the strange part. Between those two days, often decades apart, the most valuable asset in the building goes unpriced, and mostly unmanaged. Consider how a board treats the balance sheet. Quarterly scrutiny. An audit committee. External assurance, at real expense. Now consider how the same board treats the asset it calls its most valuable. An away-day, perhaps. A dashboard of media sentiment that nobody quite believes. A communications function that is invited in to describe the weather but rarely to price the risk. I don’t think this is because boards are lazy. I think it is because the asset has no number, and boards are built to govern numbers. The function that actually reads this risk gets treated as the department of words for exactly that reason: it cannot show up to the meeting with a figure. If reputation carried a price the way inventory does, the corporate affairs director would report the way the CFO does. It doesn’t, so they don’t. Now, a concession, because there is a real difficulty here and pretending otherwise would be salesmanship. Reputation is genuinely hard to measure. Every attempt I have seen to reduce it to a single score has landed somewhere between astrology and advertising, and I have watched clever people waste serious money on both. But hard to price is not the same as fine to ignore. Companies manage unpriced risks rigorously in other domains; nobody declines to run a safety programme because the value of an avoided accident is difficult to book. The absence of a number is a reason for more governance, not less. Somehow, with reputation, it has been taken as a reason for none. ## AI has started pricing continuously For most of my career, the two-pricings rule held because nobody else was doing the maths in between. That has changed, quietly and recently. Ask an AI system about a company today and it produces a settled little summary: what the firm is, what it is known for, what went wrong that time. That is a mark against your name, refreshed continuously, assembled by systems that weight what others have said about you far above what you say about yourself. It is not yet a market. But it is no longer two days a lifetime apart, either. The industry has noticed, and the response even has a name: pre-bunking. Putting material into the world in advance, so that when an AI system is asked, it finds something other than the worst version of you. The larger firms in my profession now describe doing this openly, as a service they sell. Set aside what one thinks of the practice. Notice what it admits. The most senior people in the business now operate on the basis that reputation is read daily, by AI, and that the reading can be prepared for. The years in between the two pricings, the years in which nothing was measured and nothing was managed, are being colonised by continuous assessment. Boards that found the asset convenient to ignore because it had no number are about to find it has a number every morning. ## What to do between pricings So, practically. If you own a firm you may one day sell: your eventual multiple is your reputation with a number on it, and almost everything that improves it is reputational work done years early. Making the firm’s name bigger than your own. Spreading the client base until no single departure is a story. Building the institution the buyer can actually take home. The owners who get the multiple they hoped for are the ones who understood, early, which parts of their reputation would transfer and which parts would retire with them. If you sit on a board: treat the asset the way you treat the audited ones. Assign it to someone who can be held to account. Test where it actually lives, because the crisis will tell you if you don’t find out first. Rehearse the claim before you have to make one. And when the person who reads this risk speaks, understand that the absence of a number in their hands is a measurement problem, not a seriousness problem. The two pricings are set by other people, on days of their choosing. The years in between are the only part you control. They are also, in my experience, the part almost everyone leaves unmanaged, right up until one of the two days arrives and does the pricing for them. Tim Sutton advises the owners of communications agencies on their sale, and boards in a crisis. [timsuttonpr.com](https://timsuttonpr.com) --- # What I learned buying agencies _Five essays from the seller's chair. Time to own up to what the other one taught me._ Date: 2026-08-25 URL: https://timosutton.substack.com/p/what-i-learned-buying-agencies Everything I have written here so far has been addressed, one way or another, to the founder: who really leaves in year two, what the analyst does to your claims, why the call comes a funding cycle late. In most of them I have mentioned, in passing, that with my partners I have sold a company and bought others. A few readers have asked about the second half of that sentence. Fair enough. This one is from the buyer’s chair. The received wisdom about that chair is that it is the comfortable one. The buyer has the money, the lawyers and the practice; the founder is an amateur playing a professional at the professional’s own game. I have written a version of that myself, and as a description of the negotiation it is true. But the negotiation lasts a few months. Ownership lasts as long as it lasts, and the moment the money moves, the comfort changes sides. Here is what buying actually feels like from inside. You have paid real money, today, for something you were never allowed to inspect: the private trust between your new firm’s clients and a handful of its people. Every contract you diligenced is assignable. The thing the contracts sit on is not. On completion day the seller has certainty and the buyer has hope, and I have sat in enough completion meetings to tell you which side sleeps better that night. ## The morning-after problem The strange thing about the day after completion is that nothing has changed. The clients are the same clients. The fee income is the fee income the analysts tested line by line. The people are all still on the payroll, and most of them are wearing the same expressions they wore last week. The deal has altered exactly one fact in the building: the answer to the question, whose firm is this? It took me a while to understand that this was the load-bearing answer. Clients of a founder firm are not, in the main, buying the firm. They are buying the undivided attention of a person who cannot afford to lose them, and everything else, the team, the method, the awards on the shelf, hangs off that. Change the answer to whose firm is this and every client relationship quietly goes back out to tender, whether or not anyone announces it. The buyer has not bought loyalty. The buyer has bought the right to try to earn it, at speed, from a standing start. ## The earn-out is a postponement dressed as insurance Every buyer knows the value can walk. The standard answer is the earn-out: hold back a slice of the price, tie it to the numbers, and the founder will stay and keep the clients warm. As insurance goes it looks tidy, and I have signed my share of them. But watch what the earn-out actually does to the handover. The transfer of relationships that decides whether the asset survives has usually not begun on the day the deal completes. It gets deferred into the two years afterwards, at which point it stops being a founder’s freely given endorsement and becomes a contractual obligation, performed to a formula, by somebody who has already banked most of the money and is watching the firm change around them. Both sides then wonder why the numbers drift. Neither should. The earn-out did not insure the transfer. It postponed the transfer, and charged the founder for the delay. ## Why buyers break the thing they bought The unkind explanation is stupidity, and it is wrong. The people who priced the deal understood the asset perfectly; it is why they paid so much for it. The truthful explanation is structure. The people who buy a firm and the people who own it afterwards are different people. The deal team disbands at completion. The sponsor who championed the acquisition moves on to the next one. Integration passes to managers who never read the investment case and whose year is measured in synergies, and eighteen months later one of them merges the P&L, renames the brand and moves the reporting line, because those are the things he can show for the year. The trust was in the room. The room is gone. I have used that line before, about what consolidation does at the top of the market. I should admit where it came from. It began as a note to myself, written after we had done exactly that to a firm we had chosen carefully and paid for fairly. The institutional memory of why the price was paid had left the building, and nobody still inside it was paid to remember. ## What the good buyers did differently The good ones, and I worked with some, treated the handover as the deal rather than the aftermath of it. Before completion, founder and buyer agreed, client by client, who would take each relationship on and by when, and the introductions began the week the deal closed, while the founder’s endorsement still read as a gift rather than a term of the contract. They paid the layer below the founder, properly, in the deal itself. Retention letters keep people for a notice period; a real stake in the outcome keeps them for the length of the promise. The deal insures the founder against walking, I wrote in an earlier essay and the people who hold the clients were never on the policy. The good buyers put them on it. And they left the room alone. Brand, reporting line, office, billing entity: untouched, for a stated period, until the trust had somewhere else to live. One of them kept a person on the integration whose name was on the original investment case, for no other reason than to be in the meetings where the price’s logic would otherwise be forgotten. Cheapest insurance I ever saw. ## How a founder should read all this If you are an owner heading towards a sale, this is the due diligence nobody tells you to do. Ask any buyer what happened to the last three firms they bought. Ask to speak to the founders; the good buyers offer before you ask. Count how many of the layer below stayed past the earn-out. A buyer who cannot produce one contented previous seller is telling you, plainly, that the price is the whole offer. Then weigh the offers accordingly. The biggest number attached to a careless integration is routinely worth less than a smaller number from a buyer who keeps what they buy, because most of the biggest number is contingent, and it is contingent on precisely the thing the careless buyer destroys. An earn-out hands you the integration risk without the integration authority. Price that before you admire the headline. The two chairs teach one lesson, from opposite sides. A firm like yours is a set of relationships that currently answer to one person. A sale is the project of giving them somewhere else to answer, and the project goes best when it starts before the money moves, not after. The buyers I admired never believed they were buying loyalty. They knew they were buying the founder’s help in earning it, for a period, at a price, and they behaved accordingly. They got what they paid for. The ones who believed otherwise got the laptops and the lease. Tim Sutton advises the owners of communications agencies on their sale, and boards in a crisis. [timsuttonpr.com](https://timsuttonpr.com) --- # AI has already read your crisis statement. Who did you write it for? _AI has never been better. Being believed has never been harder._ Date: 2026-08-27 URL: https://timosutton.substack.com/p/ai-has-already-read-your-crisis-statement Every panel I sit on says the same thing: artificial intelligence will speed up your crisis response. Faster drafting, faster monitoring, a first holding statement in ninety seconds instead of ninety minutes. I have heard it from boards, from agency principals, and from earnest conference stages where nobody in the room had yet been on the firing line. It is all true. And it is all beside the point, because speed was rarely the thing that lost a serious crisis. Speed, it turns out, is not even the half of it, because the machine is no longer only on your side of the table. But start with a defeat, because a defeat teaches faster than any win. ## Being right was never being believed In 1995 I was advising Shell, and the rest of the upstream oil industry, as it tried to dispose of the Brent Spar, a redundant North Sea platform, in the deep Atlantic. We had the science. Independent analysis genuinely supported deep-sea disposal as the best environmental option then available, and we led with it: charts, comparisons, expert testimony, the lot. We lost anyway, to a Greenpeace campaign and a consumer boycott that threatened to run Shell’s German forecourts dry inside a week. I have thought about that defeat more than almost any victory since. If you watch a child throw a can into the sea, you tell them it is wrong, and no measurement of comparative environmental harm touches that feeling. Ours was an argument conducted entirely in measurements. By the science of the day we were right, and being right turned out to have very little to do with being believed. Peter Melchett, who ran Greenpeace then, understood that better than any of us with our charts. He was not fighting on the facts. He had grasped that the facts were not the field. Now hand that room every tool we have built since. A modern system would have read the whole storm faster than my team could, summarised ten thousand posts before the coffee went cold, and drafted a dozen holding lines, better punctuated than mine. Not one of them would have told us the thing that actually mattered: that we had already lost the only argument in play, and that every further fact made us look deafer, not more right. ## What only a person can do That is the distinction I would put to any board, and it is the one the current excitement walks straight past. AI is extraordinary at telling you what is being said, how loudly, by whom, and which way the wind is blowing. It is hopeless at the thing that decides the outcome, which is what you should say, in your own institution’s voice, in the hour that matters, to the particular people you are talking to. The dashboard measures the weather. It cannot tell you whether to go outside. Ask what judgement actually does that the machine cannot, and you get three answers. It reads feeling, not sentiment: a score will tell you the mood is seventy-one per cent negative, but not that the crowd has quietly decided your facts are no longer the point. It chooses the messenger: it cannot look at your chief executive and know that, this time, on this issue, she is the wrong person to send out, that her very competence will read as arrogance, and that the credibility has to come from somewhere else. And it knows when to say nothing, the one move no dashboard will ever recommend, because a dashboard is built to reward activity, not restraint. The hardest advice I give is usually to say less, and occasionally to say nothing, yet. None of which slows the profession down. Deloitte’s 2026 corporate affairs report found that ninety-five per cent of corporate affairs leaders are prioritising AI upskilling, while only about a quarter have anything resembling a formal strategy for it. The industry’s own leaders now name ‘AI mastery’ as the single most critical skill they will need. When the skill a profession prizes above all others becomes fluency with a tool, it is worth asking quietly what it has agreed to stop prizing. ## The accidental automation of judgement The danger is not the tools. It is the quiet slide from using AI to think faster into using AI to decide, and those are not the same act. Look at Starbucks Korea this year, and the episode now known as Tank Day. An AI tool helped generate the wording for a promotion, and one phrase unknowingly echoed a slogan from the country’s authoritarian past. That alone would have been a bad day. What made it a crisis was duller and more familiar: several of the managers who approved the campaign had never opened the attachments they were signing off. The tool did not cause it. The absence of a single human being asking what could this mean here caused it. Joon Kim of HyperM gave the failure the name that has stuck, the accidental automation of judgement, and the mechanism is not mysterious. As Karen Yap of Kyrah & Song put it, a model optimises for the most common, safe reading of the words in front of it, and hands you the harmless one. The danger lives in the rare, the local and the historically loaded, and the machine, in her exact phrase, “doesn’t know when to be afraid”. Most reputational crises are not information problems at heart. Facts get withheld, a regulator surprises you, a document surfaces that should not have. But the damage is almost always done by a context problem, and used carelessly AI widens the information and narrows the context: more of what you already had too much of, less of the thing you were short of. And this is not a worry from before the machines arrived. When David Dao was dragged off a United Airlines flight in 2017, the monitoring worked perfectly, the footage was round the world inside the hour, and the chief executive still reached for the word “re-accommodate” and briefed his own staff that the passenger had been “disruptive and belligerent”. The dashboards were flawless. The human read was catastrophic, and the gap between the two took the share price with it. So far, so reassuring for the humans in the room. Let me put the opposite case, because I have watched advisers of my own vintage make the mirror-image mistake. It is as foolish to wave the tools away out of pride as it is to surrender judgement to them out of fashion. The adviser with forty years of instinct who says he does not need a dashboard is making his own version of the Brent Spar error, trusting the evidence he likes and ignoring the evidence he cannot feel. The answer is not to choose. It is to know exactly what each is for. ## The other side of the table And then, just as you have settled that on your own side of the table, the table changes shape. AI did not take the seat next to you. It moved into the middle. In a serious crisis now, the regulator, the journalist, the employee and the claimant each read you through a model before a single human forms a view. Four readers you did not write for, and not one of them the patient human the old playbook assumed. The regulator reads you through a model first. A supervisory team facing a tide of disclosures can run them through tools that flag the one clause that does not reconcile, or cross-reference what you say today against what you said in 2019. The Financial Conduct Authority has been plain that it will write no new AI-specific rules, leaning instead on the frameworks it already has, the Consumer Duty and the Senior Managers regime. Which means the machine does not dilute personal accountability. It sharpens it, by making the inconsistency easier to find and harder to explain away. Handing the drafting to an algorithm does not hand off the liability with it. The journalist reads you through a model first. The exclusive that once turned a story was won human to human, and that route still matters more than people think. But the first pass on your statement is now often a prompt. Summarise this. What is the company not saying? Where is the gap between the words and the filings? The reporter who used to skim your three hundred words now has a tool that lifts out the load-bearing sentence and discards the comfort wrapped around it. The wrapping was the part we used to labour over. The employee reads you through a model first. Your internal note, written with such care to reassure, is pasted into a chatbot within the hour. Decode this for me. Should I be worried. The reassurance you built for a frightened human is flattened by a tool that has no interest in being reassured, only in the proposition underneath. And the claimant reads you through a model first. This is the one that should keep a general counsel awake. A line written to sound humane, we deeply regret the distress this has caused, is, to a model trained to find admissions, an admission: distress, caused, by us. The Compensation Act 2006 says an apology is not, in itself, an admission of liability. The model has not read the Compensation Act. The claimant’s adviser no longer needs to spot the phrase. The machine spots it, flags it, and presents it as an opportunity. ## You cannot optimise for a counterparty You might reasonably object that this cuts against everything I have just argued. If machines are now the first readers, surely the answer is to write for the machine, to optimise the statement for the model the way people already optimise for search. In the ordinary run of things, courting AI as an audience is a real discipline and a sensible one. But a crisis is not the ordinary run of things. Here the model is not an audience you are trying to win. It is a counterparty reading you for the one line it can use against you. You do not get to optimise for it. You only get to be careful in front of it. And being careful in front of it is a judgement call, which is exactly the thing the first half of this argument said does not delegate. So you end up writing two documents in the same words. One is for the human, and has to land with warmth and proportion and a voice that sounds like you, because a board that sounds like a machine in its worst week loses the room. The other is the skeleton a model extracts once every comforting clause has been stripped away, and that skeleton is what actually reaches the regulator, the journalist and the claimant. The new craft is a single set of words that does both, and it is harder than the conference circuit makes it sound. It is also, I suspect, the most useful thing three decades in rooms like that one can still teach. ## The question to start asking So retire the question the industry has spent the year asking. Stop asking whether you can trust AI in a crisis. You can supervise your own use of it, you should, and that is a solved-enough problem. Start asking the one almost nobody asks: who is reading me through it, and have I written a single word for that reader? Get it right, and all that speed becomes what it should always have been, a convenience and not a strategy. Get it wrong, and you will draft the most reassuring statement of your career, and hand the claimant their best paragraph in it. The Sutton Memo is a newsletter on reputation and PR-agency deals, from someone who has spent three decades on both sides: crisis adviser to boards, and M&A adviser to agency founders and buyers. Some weeks it is the crisis. Some weeks it is the deal. [Tim Sutton PR](https://timsuttonpr.com) --- # Meta just bought the tobacco playbook _The $17bn child-safety settlement looks like a famous victory. So did tobacco's in 1998. The lesson is in what happened next._ Date: 2026-08-28 URL: https://timosutton.substack.com/p/meta-just-bought-the-tobacco-playbook In 1998 Big Tobacco lost the most famous legal battle in corporate history. [But more people smoke today than smoked then.](https://ourworldindata.org/smoking) This week Meta agreed to pay the American states [up to seventeen billion dollars ](https://www.npr.org/2026/08/26/nx-s1-5944781/meta-settlement-child-safety-lawsuit)to settle their case that it built its products to hook children. Mark Zuckerberg, who had been expected in the witness box, was spared it when the deal landed a week into the trial. The campaigners were quick to name their inspiration: the legal playbook that broke Big Tobacco in the 1990s. You can see why they reached for it. It is the most famous corporate reckoning of the modern era, and the parallel flatters everyone who worked for this result. So far, so good. But if tobacco is the model, it is worth remembering how that story actually ended, because it did not end in 1998, and it did not end well for the people it was meant to protect. ## The playbook Start with what the playbook was. The breakthrough against tobacco was not medical. Everyone had known for decades that cigarettes killed, and juries still sided with the industry, because a smoker had chosen to smoke. What changed the game was a change of plaintiff. Instead of dying individuals, the state Attorneys-General sued, and they sued not for wrongful death but to recover the Medicaid billions their taxpayers had spent treating the illness. That single move shifted the argument from personal responsibility to corporate deception, and once the industry’s own scientists and its own memoranda were in evidence, thanks to Liggett breaking ranks and to whistle-blowers like Jeffrey Wigand, the defence fell apart. The Master Settlement Agreement of 1998 bound forty-six states, cost the majors around two hundred billion dollars over twenty-five years, and killed Joe Camel. Meta’s case is that same machine, rebuilt. Forty-seven states this time, though not quite all: Florida called the deal too weak, New Mexico had already won at trial, and Texas cut its own billion-dollar deal the same day. The same theory of a company that knew. And the same figure at the centre of it: a former safety engineer, Arturo Béjar, testifying that he had warned Zuckerberg himself in 2021, and that when it came to knowing how many children under thirteen were on its products, the company’s attitude was, in his words, [“don’t ask, don’t tell”](https://www.nwaonline.com/news/2026/aug/20/kids-safety-not-priority-at-meta-bejar-testifies/). Meta denies all of it, and its settlement, in the lawyers’ customary phrase, involves no admission of liability. The litigators reaching for tobacco are not being romantic. They have built the same case, brick for brick. ## Did it work? Now the uncomfortable part, the part the retrospectives skip. Did the tobacco settlement work? Partly, and less directly than the legend allows. Smoking did fall dramatically across the West, but when economists pull the decline apart, the single biggest lever was price. The settlement, and the excise taxes that rose alongside it, made cigarettes dearer every year, and dearer is what makes smokers stop; the marketing bans mattered, but as a supporting act. And here is the largest crack in the borrowed playbook: the one remedy that demonstrably worked was price, and a free product has no price to raise. As for the money, the settlement was sold as a fund to repair the damage, and in practice the s[tates now spend about three cents in every dollar of their tobacco revenue on helping people quit](https://www.tobaccofreekids.org/what-we-do/us/statereport). The rest went where loose money always goes, into general budgets and deficit holes. A settlement, it turns out, is a transfer of cash, which is not the same thing as a cure. ## Marketing, not the product There is a deeper limit too. A settlement reforms marketing, not the product. You can retire the cartoon camel and the billboard cowboy and sell precisely the same cigarette to precisely the same lungs. Meta can cap the day at two hours, mute notifications through school time and block the app from midnight to six, and leave wholly intact the machine beneath, the one built to turn human attention into advertising revenue for as many hours as it can hold. ## Where the smoke went But the real lesson of tobacco is not in the settlement at all. It is in what the industry did the morning after. It moved. As the Western market closed around it, the business went looking for one that was still open, and found it in the developing world. The World Health Organisation reckons that around [eight in ten of the world’s 1.2 billion tobacco users now live in low- and middle-income countries](https://www.who.int/news-room/fact-sheets/detail/tobacco), and it has projected for years that by 2030 eight in ten tobacco deaths will happen there too. The treaty meant to stop this, the global tobacco convention, was signed by most of the planet. The United States signed it too, and then never ratified it. The victory was Western. The dying went south. ## Obligation, or favour Which returns us to Meta, and to the line in the settlement that ought to give the celebrations pause. The new protections are binding only in the United States. The Irish Times put it plainly the morning after: [“If you were hoping to see the same changes come to Irish teens, you might be disappointed.”](https://www.irishtimes.com/business/2026/08/27/qa-meta-settled-its-us-case-what-does-it-all-mean/) Now, Europe can largely look after itself; Britain’s Online Safety Act and Brussels’s rulebook have real teeth, and Meta will point out, fairly, [that it has already rolled out its teen accounts worldwide of its own accord](https://about.fb.com/news/2024/09/instagram-teen-accounts/). But hold that word: accord. In America, after this week, child protection is an obligation, binding, audited, priced at seventeen billion dollars. Everywhere else it is a favour, revocable by the same line of code that granted it, with no auditor and no bill attached. And optional protection is precisely what tobacco extended to the markets where eight in ten of its deaths now fall. Everywhere else is also most of the world: over half the money Meta earns, and roughly nine in every ten of the people who use its products, sit outside the United States, and the growth is in India, Indonesia and the Philippines. Tobacco’s move south took decades, ships, factories and salesmen. For software there is no journey at all. The settlement binds the smallest, richest, best-defended slice of the whole, and leaves the rest to goodwill. ## The incumbent’s move There is one further move from the old playbook, made in plain sight. The seventeen billion is not really seventeen billion. Something under thirteen is committed; the remaining four or five billion is held back, [payable only if TikTok and YouTube accept the same restrictions on the same children.](https://fortune.com/2026/08/26/meta-contingency-settlement-18-billion-tiktok-youtube/) Whatever the intent, the effect is to make the cost of child safety universal. It is the oldest instinct of the incumbent: raise the floor for everyone, and let the giant absorb what the challenger cannot. I have sat in rooms like this one, and the room is never only thinking about the charge in front of it. A settlement that looks like a defeat can be built into a moat. ## Where the parallel breaks Now let me argue against myself, because the parallel has a real weakness. Tobacco and social media are not the same kind of danger, and the difference sits exactly where it hurts the argument. By 1998 the science on cigarettes was closed. Smoking caused cancer, the industry’s own laboratories had said so, and there was nothing left to argue but the bill. The science on social media is not closed, not remotely. Careful researchers who have spent years on it, [Candice Odgers and Amy Orben among them,](https://www.nature.com/articles/d41586-024-00902-2) keep finding small and mixed effects, and warn against the confident story of a generation’s minds being rewired. Others, Jonathan Haidt the loudest, are certain the harm is grave. Addiction, in the clinical sense, is not even a recognised diagnosis here. The campaigners have borrowed tobacco’s tactics some way ahead of tobacco’s proof. There are other cracks. A cigarette has no safe dose, whatever comfort it gives, and any smoker will tell you the comfort is real. I can say so with some authority: I smoked for years, and these days I vape, which is another way of saying the nicotine and I settled out of court. That is what addiction is, a harm that feels like a benefit, and Meta’s heaviest users would say the same. The difference is the prescription. For smoking the advice is stop; nobody’s advice for a teenager is that the safe amount of Instagram is none, which is why the remedy has to be make it safer rather than make it stop, and safer is a slippery, circumventable thing. And there is a protection tobacco never enjoyed: American judges have already struck down several state laws restricting children’s social media as offences against free speech, which is part of why this ended in a cheque and not a verdict. But notice what survives all of that. The deception case never needed the academic literature to be settled. It needs a gap between what a company said in public and what it believed in private, and that gap is what the documents opened in the tobacco majors, and what Béjar’s testimony, and Frances Haugen’s before it, are opening at Meta. The science may be unsettled. The concealment is the case. So take the tobacco playbook, by all means. Just take the whole of it, including the ending. Run it forward and the forecast is not a triumph but a warning. The cheque clears. American children get their two hours and their quiet nights. The incumbents swallow a compliance cost their smaller rivals choke on. And the product, under no obligation anywhere else, carries on doing what it does, compounding quietly in Jakarta and Lagos and São Paulo, in the markets with the most children and the fewest defences. Same playbook, same ending, only faster, because this time the displacement is a setting rather than a shipping route. Last time, the world’s answer to a Western settlement was a treaty, and it took five years to write and went largely unenforced precisely where the people were poorest. This time, so far as I can see, no one has even begun to write one. In 1998 the smoke did not clear. It drifted off somewhere with weaker laws and younger lungs, and we called that a victory. Watch where this one drifts. The Sutton Memo is a newsletter on reputation and PR-agency deals, from someone who has spent three decades on both sides: crisis adviser to boards, and M&A adviser to agency founders and buyers. Some weeks it is the crisis. Some weeks it is the deal. Tim Sutton advises the owners of communications agencies on their sale, and boards in a crisis. [timsuttonpr.com](https://timsuttonpr.com) --- # Who is buying agencies now, and why it changes your number _Agency M&A is up on last year. But the buyer across the table has changed identity, and most founders are still rehearsing the old pitch._ Date: 2026-08-29 URL: https://timosutton.substack.com/p/who-is-buying-agencies-now-and-why Ask the founder of a public relations agency who will one day buy their business, and most will describe, without quite meaning to, a company that has largely stopped buying. For thirty years the answer was obvious enough that nobody examined it. You build a good independent agency, you reach a certain size and standing, and in time one of the big advertising holding companies, an Interpublic, a WPP, an Omnicom, a Publicis, comes calling. That was the exit. It was the exit I took myself in the late 1990s, when the firm I ran with my partners, Charles Barker, was sold into what became Weber Shandwick, where I went on to spend seventeen years, latterly running the business across Europe and Asia and doing a fair number of these acquisitions from the buyer’s side of the table. So I am describing a model I both sold into and later bought with. It shaped how a whole generation of founders thinks about the endgame. And it is quietly going out of date. The headline numbers look healthy enough. Data compiled by the New York law firm Davis and Gilbert, [reported by PRovoke Media at the mid-year](https://www.provokemedia.com/long-reads/article/six-first-half-agency-acquisitions-that-could-be-game-changers), shows completed deals in the public relations sector up by around a fifth on the first half of last year. The sellers are smaller. And a rising share of the buyers are private equity. [By Davis and Gilbert’s own longer count](https://www.dglaw.com/davisgilberts-13th-annual-pr-industry-trends-report-economic-uncertainty-and-ai-adoption-define-2025/), private equity and the firms it backs had climbed from about a third of completed deals in the sector to roughly half by late last year. Read those facts together and they are not really a story about volume. They are a story about who is doing the buying, and it is not who most founders are still preparing to meet. ## The buyer you were built to meet Start with the model everyone knows, because it is worth being clear about what is passing. The holding company bought for strategic reasons. It wanted a name, a client roster, a capability it had decided to own rather than build, a flag in a market it was short of. It liked to think in decades and in market share, and it could talk itself into paying a premium for all of that, because the price reflected what you were worth to its ambitions rather than what your own profit and loss could justify standing alone. You were being bought to be added to something. The founder earned out over two or three years and then, more often than not, moved on, and the agency became a line inside a larger whole. Whatever its faults, and it had many, that was the shape of the deal a generation learned to aim for. ## Why the holding companies mostly stopped So why have they gone quiet? I can tell you from inside, because I sat in the rooms as the enthusiasm drained out of it, and three things were happening at once. The first is the tyranny of the quarter. A listed holding company answers to its next set of results, and an acquisition that dilutes this year’s margin is a hard thing to defend to New York even when it is plainly right for the decade. I ran a business that had to deliver a twenty per cent margin as a matter of routine, and I know exactly how quickly a promising but dilutive deal dies in that arithmetic. The strategic decades were always a story the quarter had to approve, and the quarter grew less indulgent every year. The second is integration fatigue. The groups bought a great many agencies and discovered, expensively, how few of them were actually made better by being owned. Buying the talent turned out to be easy and keeping it worth having turned out to be hard, and after enough disappointments the appetite goes. The third is the earn-out graveyard. The very mechanism designed to hold a founder in place for three years too often purchased their quiet disengagement instead, because the number that governed their behaviour was the one they had already banked, not the one still to come. ## The dots on the map, in reverse And there is a fourth thing, the one nobody has properly written down. The old model prized breadth above almost everything, because the network was the product: an owned office in every market a client might one day mention, dots on a map that said we are wherever you need us. That philosophy built the empires, and it has quietly gone into reverse. Watch my old firm in the few years around the turn of the decade: Weber Shandwick’s offices in Australia, Ireland north and south, Malaysia, Thailand, Indonesia and Sweden were sold, one after another, to their own local managers, most of them staying inside the network as affiliates with the same brand above the door. The dot stays on the map. The ownership, the capital and the risk went home with the local management, which is to say the network became, at its edges, a franchise. My old firm was early and systematic about it, but the device is standard kit. [Omnicom used it in the same season, selling Drury, its Dublin public relations firm](https://www.provokemedia.com/latest/article/drury-porter-novelli-leaves-omnicom-in-management-buyout), back to the directors who ran it in 2020. And when the industry had to leave Russia overnight in 2022, ceding the business to local management was the route [Publicis and Dentsu both took](https://www.publicisgroupe.com/en/news/press-releases/publicis-groupe-exits-russia-while-securing-a-future-for-its-people), and one WPP named among its own. Nobody had to invent it. Some of those handovers happened on my watch, and they were the right calls, which is rather the point. A model that once paid premiums to plant flags now pays to take them down. ## The retreat reaches the centre By 2024 it had reached the centre, and you can watch it in the tape. [WPP sold its majority stake in FGS Global](https://www.wpp.com/en/news/2024/08/wpp-to-sell-its-majority-stake-in-fgs-global), one of the most prized corporate and financial names it owned, to the private equity house KKR, at a headline enterprise value of 1.7 billion dollars, and put the proceeds towards paying down its own debt. A holding company selling its best communications asset to a financial buyer to mend its balance sheet is not a freak event. It is the mood. [And the year after, Omnicom completed its purchase of Interpublic](https://investor.omc.com/news/news-details/2025/Omnicom-Completes-Acquisition-of-Interpublic-Forming-the-Worlds-Leading-Marketing-and-Sales-Company-Built-for-Intelligent-Growth-in-the-Next-Era/default.aspx), my old group, folding two of the largest holding companies into one. Consolidation on that scale is not appetite for buying agencies. It is the opposite. Within three months of completing, the merged group was combining the public relations agencies it already owned rather than shopping for new ones. When the giants are busy swallowing each other, they are not out looking for you. [And Dentsu, having tried and failed to sell its international business whole, now says it is open to deals market by market, which is the dots-on-the-map philosophy run backwards.](https://www.campaignlive.com/article/dentsu-drops-sale-international-arm-open-deals-local-markets/1948417) ## Who is buying now Into the space the holding companies have vacated has come a different kind of buyer, and it now sets the tone. Call it the platform: a communications business, usually private-equity-backed, occasionally listed in its own right, built as a vehicle to roll up others. It buys, integrates, and buys again, with a view to selling the enlarged whole to the next owner in a handful of years. ## The names are multiplying [Infinite, backed by the investment firm ParkSouth, bought Dukas Linden in New York and Greentarget in London inside seven weeks this year,](https://infiniteglobal.com/news/infinite-acquires-prominent-financial-communications-agency-dukas-linden-public-relations/) by its own account roughly doubling its weight in each city. PPHC, listed on both Nasdaq and London’s junior market, has been buying its way steadily across corporate communications and public affairs. Penta was recapitalised by Shamrock Capital last year to play the same game. Paritee, out of the Nordics, recently took a majority of an eighteen-person artificial-intelligence advisory, a firm smaller than many an account team, bought not for its size but for the capability it bolts on. Different names, one shape underneath: acquire, integrate, acquire again, and hand the whole thing to the next buyer at a higher multiple than the parts were bought for. ## They read a different number Here is why this should matter to you, and it is not a question of taste. The two buyers read different numbers off the very same business. The holding company was buying a strategic position, and could persuade itself to pay a premium for it. The platform is buying earnings and a multiple, and it is buying them in order to sell them on. That difference changes three things a founder ought to care about. It changes the price, because the sum now turns on what your profit genuinely is, cleanly and defensibly, rather than on the story of what you might one day be worth to somebody’s grand design. It changes the clock, because a sponsor has a fund life and bought you intending to exit, so your earn-out is not a gentle handover into a permanent home but one leg of another party’s return, and your remaining equity may be sold, and sold again, across the very years you are still working it out. And it changes what the word fit means. The holding company wanted a jewel. The platform wants a component that slots in cleanly and lifts the group’s own multiple, which means that legibility beats brilliance, and the very idiosyncrasies that made you distinctive can read, to this buyer, as integration risk. ## The exception worth naming There is always an exception, and this one is worth naming precisely because it does not undo the point. Havas has spent the year building rather than shrinking, taking a majority of the Spanish public affairs firm Acento and buying the French influence consultancy Format, folding both into its H/advisors corporate advisory arm. It is a holding company moving deliberately against the mood, at the high-margin, high-stakes end of the business. But one builder among several sellers is not a counter-trend, it is a niche, and a revealing one, because even the exception is chasing the same corporate and public affairs work the platforms most want. The prize everyone is converging on is the advisory end, where the margins are, and the commoditised middle is left to fend for itself. ## Dress for the buyer you have None of this is cause for despair, and it is certainly not a reason to sell in a hurry into a market that rewards the patient. It is a reason to know which buyer you are dressing the business for, because the two want different things and reward different preparation. Ready yourself for the holding company that once defined the exit, and you may find you are polishing a story of strategic value for a buyer who has stopped paying for it. Ready yourself for the platform, and the work is duller and far more valuable: clean and legible earnings that survive a stranger’s scrutiny, a business that runs without you in the room, and an honest view of what your profit is actually worth to someone who means to sell it on. The buyer changed while much of the industry was still rehearsing the old pitch. The founders who do best over the next few years will simply be the ones who noticed. The exit you spent a career imagining may well still exist. Just make sure it is the one still writing the cheques. The Sutton Memo is a newsletter on reputation and PR-agency deals, from someone who has spent three decades on both sides: crisis adviser to boards, and M&A adviser to agency founders and buyers. Some weeks it is the crisis. Some weeks it is the deal. Tim Sutton advises the owners of communications agencies on their sale, and boards in a crisis. [timsuttonpr.com](https://timsuttonpr.com) --- # The archive is the character witness _Your next crisis statement will be read against ten years of your own words, by machines, in seconds. The verdict lands before your first call connects._ Date: 2026-08-30 URL: https://timosutton.substack.com/p/the-archive-is-the-character-witness The first reader of your next crisis statement will not be a journalist, a customer or a regulator. It will be a machine. And it will not read the statement on its own terms, the way a press officer reads it, checking that each sentence is defensible. It will read it against everything you and your organisation have published in the last ten years, and it will form a view, in seconds, on whether the two rhyme. That view is not a curiosity. It sits in the summary box above the coverage, it shapes the first question from the first journalist, and it is quietly read by your own directors on the way to the emergency board call. So here is a question worth sitting with in the calm, while it is still cheap to answer: how much of your last decade would you like read back to you tonight? I will give you the long answer, because the short one is unkind. ## The witness that has read everything In court, a character witness testifies to your record: not to the incident, but to whether the incident sounds like you. Reputation has always had character witnesses. They used to be human, slow and few. The correspondent who had covered you for twenty years. The analyst with a long memory, the rival with a longer one. Their testimony took days to assemble and reached a limited audience. That role has now been automated. Ask any of the AI engines about a company in trouble and it does not answer with the event alone. It answers with the event set against the record. What did they say about safety before the recall? How does the apology sit beside the values page, the chairman’s letters, the old boasts to the trade press? The comparison that once took an investigative reporter a week in the cuttings library is now the default shape of every answer, produced for anyone, at no cost, before your statement has finished its approvals cycle. The character witness, in other words, has already been briefed. Ten years of your own words are the briefing. And unlike the human version, this witness is available to everyone, at any hour, and testifies without being called. ## What cross-examination used to cost The record always mattered, so let me be fair to the received view. Consistency has been part of the reputation trade forever, and nobody senior needed a machine to tell them that today’s statement should not contradict last year’s annual report. But checking was expensive. It took opposition researchers, discovery lawyers, a well-funded newsroom, or a grudge with stamina. Most inconsistency simply died in the archive, unread, because reading was labour. That cost has gone to zero, and zero changes behaviour. Nor is it just the crisis query. Every routine question about you (is this firm any good, who runs it, what do they stand for) is now answered from the record, by a system that has read all of it and forgets none of it. The machine is not a filing cabinet of what happened. It is an opinion, already formed, drawn from what you said, refreshed every time somebody asks. Which would be manageable, if the opinion were reliably accurate. ## I asked it about myself In July I ran a blind test on my own name: a set of questions put to the engines in clean, logged-out browsers, worded the way a buyer or a board member would word them, not the way I would like them worded. The witness knew my record disconcertingly well. Mandates from twenty-five and thirty years ago came back accurately summarised, work I had half-assumed was beyond anything a machine could still reach. If you have spent decades in public life in any form, the archive is deeper than you think, and all of it is in evidence. And the witness misremembered me anyway. Every engine I tested described a chairmanship I had stepped back from as my current job; one put it in the opening sentence, present tense, as the headline fact of my career. Another wove a different Tim Sutton’s career into mine, mid-answer, without flagging the join. Nothing scandalous, nothing hostile. Just wrong, confidently, in the exact places where a stranger forming a first impression would look. Fixing it was instructive. There was nobody to ring. You cannot cross-examine a summary box, and you cannot send it a correction. The only lever was the record itself: publish the current fact, dated, at the source the machines actually read, and wait for the next crawl. It took an afternoon, and it was only possible because nothing was on fire. Whether the witness has updated its briefing, I will find out the way everyone does: by asking it again. ## Wrong, and still deciding The obvious objection, and it is a fair one: these systems are confidently sloppy (I have just told you so myself), no serious person takes the summary box as gospel, and human beings still make the actual decisions. Well, yes. To a point. But the verdict does not need to be right to be operative. The machine’s reading arrives first, and first framings harden. I have spent much of my working life inside the opening hour of corporate crises, and the whole discipline of that hour was arriving before the story set. The story now begins to set before your first call connects. The junior reporter’s background paragraph, the analyst’s briefing note, the questions your own non-executives bring to the call: increasingly, all of it is drafted from the same machine summary, produced in the seconds after the news broke, judged against your decade of words. And if the witness misremembers, that is an argument for deposing it early, not for hoping it will not be called. A wrong briefing you discover in the calm is a correction. A wrong briefing you discover in a journalist’s question at nine in the evening is a second crisis, running alongside the first, of exactly the kind nobody has spare hands for. ## Signed, or merely issued Now the newer fact, the one I have been sitting on for a few weeks waiting for its moment. Since the second of August, under the [transparency provisions of the EU’s AI Act](https://artificialintelligenceact.eu/article/50/), Anthropic has [begun embedding an invisible watermark](https://www.anthropic.com/news/claude-text-watermark) in the text its Claude models produce. Google has [marked Gemini’s text output since 2024](https://deepmind.google/blog/watermarking-ai-generated-text-and-video-with-synthid/). There is, so far, no public detector for text: the marks are being laid down now, and the tools to read them are [promised but not shipped](https://support.claude.com/en/articles/16266773-how-claude-marks-ai-generated-content). The direction of travel, though, is not in doubt. At some point, and I would not bet on it being far off, it will be possible to put a new question to a crisis statement: did a machine have a hand in this? Here is the nuance that will be missed in the first hundred headlines about it. The watermark means a machine processed the words, not that it wrote them. Run a chief executive’s heartfelt, hand-written apology through a machine to check the spelling, and it may carry the same mark as a statement no human ever touched. So the scan, when it arrives, will settle almost nothing. And it will be run anyway, by everyone, because a positive result makes an irresistible story: the apology nobody wrote. What answers that story is not a counter-scan. It is the record. A decade of words in a recognisable voice, dated, much of it laid down before these tools existed, that sounds like the person now standing at the microphone. Authorship, it turns out, is about to become part of character, and the archive is the only witness to it. My own position has been settled, in writing, for a while. Every word published under my name is reviewed, rewritten where needed, and signed off by me, with a date on the sign-off. The machines I use daily in the drafting never get the final word. I would rather own those sentences in the calm than improvise them into a hostile microphone. ## What to do in the calm The practical part, and none of it requires a consultant, which is rather the point. First, depose the witness. Ask the engines about yourself, your firm, your board, in a clean browser, logged out, in the words a sceptical stranger would use. Not once: quarterly. You are not checking your publicity. You are reading the briefing that will frame your worst night. Second, read your own decade the way the witness does. Somewhere in it is the sentence that will be set against your next statement: the values page written in better times, the confident claim in an old interview, the boast that has quietly become a liability. Find it before it is found. Third, correct in peacetime, at source. Stale and wrong facts get fixed where the machines read, with dates on the fix, while nothing is burning. Midnight is for the crisis you have. It is a bad hour to start amending the record you need. Fourth, decide your authorship line before you are asked. Who writes your public words, what part machines play, and what you will say when the question comes. It is coming for everyone. The organisations that will look worst are not the ones using the tools; they are the ones that never decided what to say about it. Under a note of mine last week, a reader named Robin Dimond compressed all of this into one line better than I had managed in several paragraphs: “You can’t manufacture authority at the moment you need it.” Quite. And the record is where the machines now go looking for the authority. Your statement can be drafted tonight. The briefing it will be judged against was drafted over the last ten years, by you, one unguarded paragraph at a time. The question I put to boards on the worst night of their year is the one worth putting to the machine while the year is still quiet: what are you most hoping no one asks? Ask it now, while the answer is still something you can do something about. The Sutton Memo is a newsletter on reputation and PR-agency deals, from someone who has spent three decades on both sides: crisis adviser to boards, and M&A adviser to agency founders and buyers. Some weeks it is the crisis. Some weeks it is the deal. Tim Sutton advises the owners of communications agencies on their sale, and boards in a crisis. [timsuttonpr.com](https://timsuttonpr.com) --- # The fuse in the lunchbox _American lawyers are re-running the tobacco playbook against the biggest names in food. Britain has not noticed yet _ Date: 2026-09-01 URL: https://timosutton.substack.com/p/the-fuse-in-the-lunchbox Everyone knows about ultra-processed food now. The acronym has travelled from the nutrition journals to the school gate in about five years, Chris van Tulleken’s book sat on the bestseller lists for months, and there cannot be a food marketer left in Britain who has not sat through a slide about it. The received view is that this is a health story: a matter for dietitians, food writers and the more energetic sort of parent. Well, yes. To a point. ## It starts on a court docket Because the development that will matter most did not appear in a medical journal. It appeared on a court docket. In December 2024 a writ landed in Philadelphia on behalf of a young man diagnosed with type 2 diabetes and fatty liver disease as a teenager, naming eleven food manufacturers. In December 2025 the City Attorney of San Francisco [sued a roster of the biggest names in packaged food](https://sfcityattorney.org/san-francisco-city-attorney-chiu-sues-largest-manufacturers-of-ultra-processed-foods/). In April 2026 a [class action in Wisconsin](https://www.fooddive.com/news/ultraprocessed-food-consumer-lawsuit-kraft-heinz-pepsico/818951/) named twelve defendants and asked for a billion dollars. A steady run of consumer suits has followed through this year. The defendants read like the shelf nearest the till: Kraft Heinz, PepsiCo, Coca-Cola, Nestle, Mondelez and, in the Wisconsin case, Unilever. ## The playbook, by name The allegations, and they are allegations, are that these products were formulated so that people would eat more of them than they intend to, and that they were marketed to children in that knowledge. The word the pleadings keep reaching for is addiction. And the drafters are not remotely coy about their model. These cases are built, clause by clause, on the tobacco litigation of the 1990s. The pleadings invoke the tobacco playbook by name. I wrote here recently about the same playbook turning up somewhere else entirely, in [the litigation that trailed Meta’s settlement with the states over child social-media addiction](https://timosutton.substack.com/p/meta-just-bought-the-tobacco-playbook). The charge was the same one being made here: a product engineered to hook the young, and marketed to them anyway. Same script, a different industry. Seeing it reach the food aisle is seeing a pattern, not a coincidence. ## The forty-eight-year fuse It is worth being precise about what actually happened with tobacco, because the folk memory has compressed it. The science was not a late discovery. Doll and Hill [published the link between smoking and lung cancer](https://csts.ua.edu/files/2019/01/1950-09-30-BMJ-Smoking-Carcinoma-of-the-Lung-Doll-Hill.pdf) in the British Medical Journal in 1950. The industry’s own researchers confirmed it internally within years. A [Brown & Williamson memo of 1969](https://www.sourcewatch.org/index.php/Smoking_and_Health_Proposal) stated the strategy in four words that should be taught in every business school: “Doubt is our product.” In 1994 seven chief executives stood before Congress and testified, under oath, one after another, that they believed nicotine was not addictive. In 1998 the industry signed the [Master Settlement Agreement](https://www.naag.org/our-work/naag-center-for-tobacco-and-public-health/the-master-settlement-agreement/): 206 billion dollars. Forty-eight years from the science to the settlement. Notice the shape of that. Everything that mattered was in the public record for decades before it was believed. A crisis is not really an event at all. It is the moment when the gap between what the record shows and what the public believes finally closes, and it closes all at once. ## The prize is discovery Now, the food cases may fail. The first of them already has: the Philadelphia suit was [thrown out in August 2025](https://www.foodbusinessnews.net/articles/30636-court-tosses-upf-lawsuit-against-food-companies) for failing to tie particular products to the plaintiff’s illness, and in June this year a federal judge dismissed the amended version and refused to allow another. Causation in diet is genuinely harder than causation in smoking; the confounders are everywhere, and no sensible lawyer expects a settlement soon. But the first tobacco cases failed too. They failed for forty years, and the food suits are [multiplying regardless](https://news.bloomberglaw.com/health-law-and-business/ultra-processed-food-lawsuits-multiply-despite-early-setbacks). Watch San Francisco in particular: a government plaintiff suing on public nuisance is the vehicle that finally cracked tobacco, and opioids after it, because it is the route by which internal documents reach the record. What destroyed the tobacco industry’s position was not epidemiology. It was its own filing cabinets. ## An American eccentricity? And Britain? No writ here. The government [scrapped its planned ban on junk-food multibuy deals](https://www.foodnavigator.com/Article/2025/07/09/hfss-promotions-ban-scrapped-in-the-uk/) in July 2025, weeks before it was due to take effect, preferring what it calls partnership with industry: reformulation and education rather than statute. From a London boardroom the whole thing can be made to look like an American eccentricity, like litigation about coffee temperature. Yet look at what the companies are doing rather than what anyone is saying. Across the big portfolios, sugar and salt are coming down. Recipes are being reworked, quietly, product by product, with no dates attached and no speeches given. Here is a question worth sitting with: if the boards of these companies genuinely believed the science would come to nothing, why spend margin on reformulation? Quiet reformulation is what discounting the future looks like when nobody wants to say so out loud. ## Another way to play it There is another way to play a lit fuse, and I can describe it with some confidence, because I helped make the move. By the end of the 1980s, alcohol was where food is now: the science hardening, the tabloids full of lager louts, and ministers openly reaching for statute. The producers could have reformulated quietly and hoped. Instead they came together and built the [Portman Group](https://www.portmangroup.org.uk/about-our-history/): their own watchdog, funded by the industry. Launching it was my brief, on behalf of the combined drinks coalition, and the appointment at the top told you everything about the intent: Dr John Rae, former headmaster of Westminster School. Nobody hires a headmaster unless they expect to be told off. The Group had to be seen to police its members, and it was: companies were ordered to change or withdraw advertising that overstepped, and young drinkers got the UK’s first proof-of-age card. When the alcopops panic arrived in the mid-90s, with fruit-flavoured drinks the tabloids said were dressed up for children, the Group answered with its 1996 code on how drinks could be named, packaged and promoted, and gave it teeth: a product found in breach goes on an alert to the retailers, who agree not to restock it. Three decades on, alcohol marketing in Britain is still substantially self-regulated. The statute never came. The move was loud, collective and early, and that is the entire point. It cost the producers something at the time: money, pride, and the discomfort of admitting in public that the product carried risks worth managing. What it bought was thirty years of holding their own pen. ## Before the gap closes Managing the day the gap closes has been a fair part of my living, and I will let you in on a trade secret: the advice is far better value before it does. The latent phase, the years when the fuse is visibly lit but the building has not caught, is the only period in which a board has real choices. Change quietly, and this quarter is protected; but in a courtroom in 2031, the date you started reformulating becomes the date you knew. Change loudly, with dates and numbers attached, and it costs you now, in margin and in awkward headlines, but the record you build becomes your defence. I have written here before that [the archive is the character witness](https://timosutton.substack.com/p/the-archive-is-the-character-witness). This is what I meant. The archive these companies are writing right now, in their product decisions and their silences, is the one they will be cross-examined on. ## There is always a defector One last thing from the tobacco files. There was, eventually, a defector. Liggett, the smallest of the majors, broke ranks in 1996 and settled first; within a year it had admitted what the industry had spent four decades denying and handed over its documents. It got terms the giants never saw. Every tobacco-shaped story produces a Liggett in the end, and the economics of defection reward whoever moves while it still looks like a choice. Somewhere inside one of those twelve defendant companies, somebody is already drafting the memo that says: go first. History suggests the board will ignore the first draft. It also suggests the company that eventually listens will not be the biggest in the room. Which one would you back? The Sutton Memo is a newsletter on reputation and PR-agency deals, from someone who has spent three decades on both sides: crisis adviser to boards, and M&A adviser to agency founders and buyers. Some weeks it is the crisis. Some weeks it is the deal. Tim Sutton advises the owners of communications agencies on their sale, and boards in a crisis. [timsuttonpr.com](https://timsuttonpr.com) --- # Where the water went _Big Tech promised to be water positive by 2030, with dates attached. The planning system is now publishing the real numbers one town at a time. Britain's answer is to stop the towns asking._ Date: 2026-09-05 URL: https://timosutton.substack.com/p/where-the-water-went Last August, in the driest six months England had recorded since 1976, the National Drought Group [published its advice to households](https://www.gov.uk/government/news/national-drought-group-meets-to-address-nationally-significant-water-shortfall). Shorter showers. Fix the dripping tap. And, on the same list, this: “Delete old emails and pictures as data centres require vast amounts of water to cool their systems.” The country laughed, the trade press winced, and within a week the line had gone round the world as the moment Whitehall blamed the drought on your inbox. It was a silly sentence. It was also, I think, the most honest thing any official body in Britain has yet said on the subject, because it let slip a connection that the people building the data centres have spent five years and a good deal of money keeping apart: the water in the reservoir and the machine in the shed. The received view is that data centres are an energy story, and a specialist one at that. Gigawatts, grid queues, something for the infrastructure pages and the people who read them. Well, yes. To a point. ## It starts with a planning application When I wrote here about [the tobacco playbook](https://timosutton.substack.com/p/the-fuse-in-the-lunchbox), I argued that a crisis is the moment the gap between what the public record shows and what the public believes closes all at once, and that the cigarette companies kept that gap open for forty years because nobody could force the documents out. Discovery, when it finally came, came through the courts, one subpoena at a time. Data centres are the opposite case, and it is the most interesting thing about them. Discovery is built into the product. Nobody can build one without filing a planning application, and a planning application states, in black and white, how much power the thing will draw and how much water it will drink. Every local fight over a data centre is therefore a discovery process that the industry did not have to hand over. The residents of a small town in Oregon or Arizona are doing, for nothing, the work the plaintiffs’ bar spent a generation and a fortune doing to Philip Morris. Consider The Dalles, Oregon. In 2021 the state’s biggest newspaper asked the city how much water Google’s data centres there were using. The city refused, called the figure Google’s trade secret, and sued the newspaper to keep it that way, with Google paying more than $100,000 towards the city’s legal bill. When [the case settled](https://www.rcfp.org/dalles-google-oregonian-settlement/) in December 2022 the number came out: 355 million gallons in 2021, 29 per cent of everything the town used. Google then announced it would no longer treat site-level water use as a trade secret anywhere. Note the order of events. The disclosure did not come from the pledge. It came from the fight. Consider Tucson, where a proposal arrived in 2025 under a code name, Project Blue, with officials bound by non-disclosure agreements about who was behind it. The council [voted seven to nil to reject it](https://www.democracynow.org/2025/8/7/headlines/tucson_city_council_votes_unanimously_to_reject_project_blue_a_proposed_data_center_linked_to_amazon) in August 2025; the backer turned out to be linked to Amazon, which walked away. The same story is running in Virginia, the Netherlands, Chile and Uruguay; in Ireland, where data centres take 23 per cent of all metered electricity and the grid operator has stopped connecting new ones in Dublin until 2028; and in Aragon, where the campaign against Amazon carries a name that translates as Your cloud is drying my river. Data Center Watch counted [$64 billion of American projects blocked or delayed](https://www.datacenterwatch.org/report) by local opposition by March 2025, and noted that the opposition was bipartisan. By its count the first half of this year alone passed $200 billion, and in July New York enacted [the first state-level moratorium](https://www.rockinst.org/blog/updates-on-the-cloud-more-moratoriums-on-data-centers/). This is not a fringe. It is the fastest-growing planning phenomenon in the western world, and it runs on the industry’s own paperwork. ## The pledges, with their dates Now sit the pledge layer on top of that, because this is where a planning fight turns into a reputational one. Between January 2020 and November 2022 the four largest builders of data centres made a set of promises unusual for their precision. Microsoft would be carbon negative by 2030 and water positive by 2030. Meta: water positive by 2030. Google: replenish 120 per cent of the water it consumes by 2030. Amazon: water positive by 2030. Dates, percentages, baselines. It was exactly what a good adviser would have told them to do in 2020, and it is what I would still tell them to do. A pledge with a date attached is a hostage to the record. That is the point of it. It is also the risk. Then the arithmetic ran the other way. ## The arithmetic ran the other way I have taken what follows from the companies’ own reports, not from their critics. Microsoft’s [2026 environmental data fact sheet](https://cdn-dynmedia-1.microsoft.com/is/content/microsoftcorp/microsoft/msc/documents/presentations/CSR/2026-Microsoft-Environmental-Data-Fact-Sheet-PDF.pdf), reviewed by Deloitte, gives total emissions of 13.06 million tonnes in the 2020 financial year and 21.12 million in 2025: 62 per cent above the baseline year of the carbon-negative pledge, and up by a quarter in a single year. Water consumption went from 3,990 megalitres to 8,170 over the same period, more than double what it was when the water pledge was made. Microsoft also reports that it replenished more than it withdrew last year, which is creditable and does not change the direction of the line. Google’s total emissions were 51 per cent above its 2019 baseline by last year’s report, and its electricity demand rose 37 per cent in 2025 alone. In June [its own report](https://blog.google/company-news/outreach-and-initiatives/sustainability/2026-environmental-report/) said: “reaching our climate moonshots is getting harder”, and then, in a sentence I would frame: “Our AI infrastructure buildout is accelerating faster than the grid is decarbonizing.” Amazon’s rose 16 per cent in 2025. Put the three together, [as the Guardian did in July](https://www.theguardian.com/us-news/2026/jul/11/microsoft-amazon-google-datacentre-carbon-emissions-france), and you get 119 million tonnes in the latest year, up nearly a fifth, or about a third of France. All three still hold to their net-zero dates. None of this is hidden. Every number in the last two paragraphs is public, audited and voluntarily published. That is precisely what makes it a latent crisis rather than a scandal. The record is complete. The belief has not caught up with it. ## A fifteenth of a teaspoon And here the industry did something I recognised at once from an earlier life. It reached for a reassuring unit. In June 2025 Sam Altman [wrote](https://blog.samaltman.com/the-gentle-singularity) that the average ChatGPT query uses “about 0.000085 gallons of water; roughly one fifteenth of a teaspoon”. Two months later Google published a paper putting a median Gemini prompt at 0.26 millilitres of water and 0.24 watt-hours. Amazon’s latest report says its data centres are “seven times more water-efficient than the industry average” and that in 2025 it “returned three gallons for every four it used”. Each of these figures may be perfectly true. Academics, Shaolei Ren at the University of California, Riverside, among them, pointed out that the teaspoon leaves out the water used to generate the electricity, which in most places is the larger share. But the deeper objection is simpler. Efficiency per query is a ratio. A reservoir is a quantity. You can be seven times more efficient than average and still take a quarter of a town’s water, if you are big enough, and being big enough is the business plan. The trick is an old one: change the unit until the number sounds like nothing. A fifteenth of a teaspoon is a fine answer to the question, how much water does my query use? It is no answer at all to the question the people of The Dalles were asking, which was, how much of our water are you using? The two questions have different units, and only one of them appears on a planning application. Nor will any unit beat the intuition on the other side, which is a hosepipe ban on your garden while a windowless building a mile away hums along on the same mains. All the arithmetic in the world loses to that. The only thing that has ever beaten it is having been seen to say so first. What settles it, for me, is the memo. In October 2025 SourceMaterial and the Guardian published [a leaked 2022 AWS water strategy document](https://www.source-material.org/amazon-leak-reveals-true-data-centres-water-usage-secret-plan/) in which the company chose to count only the water it used directly, not the water used to make its electricity, “because of reputational risk”; noted that counting both would roughly double the figure; called the release of water data “a one-way door” to be opened only if regulators insisted; and predicted the headline it feared, “Amazon hides its water consumption”. Amazon says the memo is obsolete. Perhaps it is. But it is in the record now, and if the tobacco years taught the reputation trade anything, it is that the memos matter more than the science. Not because they prove harm. Because they prove knowledge. ## Britain stores the gap Which brings me to what Britain has done, and it is, from where I sit, the most dangerous option on the table. In September 2024 the government designated data centres as Critical National Infrastructure, and the phrasing was its own: [“alongside energy and water systems”](https://www.gov.uk/government/news/data-centres-to-be-given-massive-boost-and-protections-from-cyber-criminals-and-it-blackouts). When Three Rivers District Council refused a data centre on green belt at Abbots Langley, the Secretary of State called the decision in and approved it. The first AI Growth Zone went to Culham in Oxfordshire, in Thames Water’s supply area, which the Environment Agency classes as seriously water stressed, seven miles from where the company hopes to build a reservoir it does not yet have. Global Action Plan [found this April](https://www.globalactionplan.org.uk/insights/news/report-reveals-majority-of-new-data-centres-planned-for-water-stressed-areas) that 84 per cent of the water-intensive data centres proposed in Britain sit in areas already water stressed or forecast to be by 2040. Thames Water says a large facility might use “anywhere between four and 19 million litres of water per day”. The Environment Agency’s [national water framework](https://www.gov.uk/government/publications/national-framework-for-water-resources-2025-water-for-growth-nature-and-a-resilient-future/9-taking-action-on-other-significant-water-using-sectors-and-emerging-demands-national-framework-for-water-resources-2025) notes “the potential for a new large demand as more data centres are built” and suggests the centres look for their water somewhere other than the mains. There are no official figures for how much water British data centres use. When The Times asked, half the water companies could not say how much they supply. And when the Green Party’s leader called last week for a pause on new approvals ([“slam the brakes on these energy guzzling, water guzzling data centres”](https://greenparty.org.uk/2026/08/28/green-party-calls-for-govt-to-slam-the-brakes-on-new-data-centre-approval-after-water-shortage-fears/)), the government replied within the day that a pause would be “a disaster for jobs and national security”. He may be wrong about the pause and they may be right about the jobs. That is not my point. My point is about the gap. Taking data centres out of local hands does not close the distance between the record and the belief. It stores it. The American towns are closing theirs one planning committee at a time, noisily and expensively, where people can see the numbers. Britain has arranged for its numbers to come out nationally, all at once, in a drought summer. And the drought summer is this one. On 29 July the Environment Agency [declared seven regions in drought](https://www.gov.uk/government/news/drought-declared-in-half-of-england), including the whole of London and the Thames Valley, with reservoirs at 75 per cent and 23 million people under hosepipe bans. When I sketched this essay at the start of the month I gave the fuse a year or two. I now think it is shorter. ## Loudly or quietly The board lesson from tobacco transfers here without bending, and two of the companies have already acted out the two halves of it. Change loudly, with dates attached, and the record becomes your defence. In May 2024 Brad Smith, Microsoft’s president, [told Bloomberg](https://www.bloomberg.com/news/articles/2024-05-23/a-big-bet-on-ai-is-putting-microsoft-s-climate-targets-at-risk) that because of AI “the moon is five times as far away as it was in 2020”. Nobody enjoyed hearing it. But it was said, on the record, and this summer Microsoft began [publishing water and electricity figures for individual data-centre regions](https://blogs.microsoft.com/blog/2026/06/24/inside-microsofts-two-decade-push-to-cut-water-intensity-while-scaling-for-growth/). You can dislike the numbers. You cannot say they were hidden, and when the reservoir is at 75 per cent, that is the sentence that matters. Change quietly, and the date you started becomes the date you knew. In 2024 Google stopped claiming to be carbon neutral, a claim it had made since 2007. There was no announcement. The change appeared in the report: “Starting in 2023, we’re no longer maintaining operational carbon neutrality.” [Bloomberg noticed](https://www.bloomberg.com/news/articles/2024-07-08/google-is-no-longer-claiming-to-be-carbon-neutral). It was the right decision, since the offsets were poor, taken in the wrong way, and it sits in the record now as a retreat rather than a correction. For the builders the lesson is short: put the water and power in the application before the objectors do, in the units the town uses. For the far larger number of boards that use these companies rather than build them, there is a second lesson, and it is the one I would spend a board’s afternoon on. Your supplier’s pledge is your pledge. The FTSE company whose net-zero claim rests on a cloud contract is carrying emissions and water in somebody else’s shed, on a promise the shed’s owner has already said in print is getting harder to keep. Saying that you use Azure does not launder that. It only means that when the gap closes you will be reading about it rather than writing it. ## There is always a defector In the tobacco story it was Liggett, the smallest of the majors, that broke ranks in 1996 and handed over the documents. There is always a defector, because the reward for being first to tell the truth grows as the gap closes. The first ‘defector’ here has already appeared, and he is not a company. He is a whistleblower. In July a former AWS water sustainability programme manager, Dr Nathan Wangusi, [filed suit in Arlington County, Virginia](https://www.theregister.com/on-prem/2026/07/15/aws-sustainability-claims-dont-hold-water-lawsuit-alleges/5269723), alleging that the company’s public claims about its Northern Virginia data centres are false: among them that they run “ninety-seven percent of the year by pulling outside air and not using any water”, and that AWS is 75 per cent of the way to water positive. His evidence is not a leak. It is the water bills, obtained under freedom of information law from the local water authority, which he says show withdrawals in every month of the year, winter included. These are allegations; Amazon says its 2025 figures were assured by a third party, and a court will decide. But look at the shape of the case. The record he relies on belongs to the utility, and it was public all along. Where does this end? Not, I suspect, in a Master Settlement Agreement. The likelier shape is the one Ireland has already reached without anybody calling it a crisis: the grid operator stopped connecting data centres in Dublin, and [the regulator now makes new ones bring their own supply](https://www.cru.ie/about-us/news/the-cru-publishes-its-decision-on-new-electricity-connection-policy-for-data-centres/). Utilities are the ultimate defectors. They cannot be spun, and they have to answer the phone when the taps run dry. For the companies the choice is the one it always was. Publish the per-site water and power now, in the application, in the units the town uses, and have the argument in litres rather than teaspoons. Or keep the door one-way, and let the drought open it for you. Delete your old emails if it makes you feel better. It will not save the reservoir. But it might remind you where the water went. The Sutton Memo is a newsletter on reputation and PR-agency deals, from someone who has spent three decades on both sides: crisis adviser to boards, and M&A adviser to agency founders and buyers. Some weeks it is the crisis. Some weeks it is the deal. Tim Sutton advises the owners of communications agencies on their sale, and boards in a crisis. [timsuttonpr.com](https://timsuttonpr.com) --- # Being right - and still being beaten _Winning the fact, losing the room. What Shell taught me in 1995, and why boards keep making the same mistake._ Date: 2026-09-07 URL: https://timosutton.substack.com/p/being-right-and-still-being-beaten Last April the Advertising Standards Authority looked at a Shell advertisement, weighed the complaints against it, and [cleared it](https://www.asa.org.uk/rulings/shell-uk-ltd-g24-1248246-shell-uk-ltd.html). That October it [did the same again](https://www.asa.org.uk/rulings/shell-energy-uk-a25-1288971-shell-energy-uk-ltd.html). On the first of those two days it also [banned an advertisement by TotalEnergies](https://www.asa.org.uk/rulings/totalenergies-se-a24-1246357-totalenergies-se.html). A ruling like that settles one narrow thing, whether a particular advert broke a particular rule. It is not a verdict on whether the company can be trusted, and the public has never confused the two. You would not guess, from the way Shell is talked about, that the regulator keeps finding in its favour. There is a comfortable belief in my trade, and in most boardrooms, that this is how you win. Get the facts on your side, marshal them well, and the argument follows. Be right, and be seen to be right. Well, yes. To a point. Because a company can be right about every number, get its fiercest critic to admit in writing that it was wrong, have an independent audit say the same, and still lose. Completely. I know, because I watched it happen, and the first time I saw it the company was also Shell. ## Two arguments at once It is worth being precise about the mistake, because it is not the obvious one. The mistake is not caring too much about the facts. It is leading with them, into an argument that was never about them. Every serious reputational fight has two arguments running at once. There is the argument on the table, the one with the numbers and the engineering and the technical case. And there is the argument in the room, about what kind of company this is, whose side it is on, whether it can be trusted. The first is the one you can win. The second is the one that decides the outcome. Confuse the two, answer the second with the ammunition meant for the first, and you can be entirely, demonstrably right and finished by lunchtime. ## The child and the Coke can In 1995 I was advising the offshore oil industry, which is a polite way of saying I had a ringside seat at the worst reputational rout of the decade. Shell had a redundant oil storage buoy in the North Sea called the Brent Spar. It had reached the end of its life, and the company proposed to sink it in deep water in the North Atlantic, more than two kilometres down. This was not a rogue plan. It had been studied, it was the option Shell’s own analysis called the best practicable environmental one, and the British government had approved it. Greenpeace occupied the platform on the last day of April and turned it into a symbol. In Germany, Shell service stations were boycotted and some were attacked, and sales fell steeply, by some accounts by around a half. The figure that has clung to the story ever since, that the Spar still held more than five thousand tonnes of oil, came late, in the middle of June, days before Shell caved and six weeks into a campaign that was already winning without it. On the twentieth, Shell gave up and [abandoned the deep-sea plan](https://www.ogj.com/home/article/17217041/shell-abandons-plan-to-dump-brent-loading-spar-in-atlantic-ocean). The case against Shell needed no expertise at all, and that was its power. Walk along a beach, watch a child throw an empty Coke can into the sea, and you would tell them at once that it was a naughty thing to do. Everyone knows the sea is not a bin. So if a child may not toss a can into it, how can a company be allowed to sink thousands of tonnes of redundant steel and oil into it instead? That was the whole argument, and it fitted on a placard. Shell answered it with a different one. Its case was arithmetic. Set against the Atlantic, the oil left in the Spar was trivial, a can of pop in an ocean, and the deep sea bed was genuinely the least harmful place to leave the structure. On the science this was defensible: [a Nature editorial said as much](https://www.nature.com/articles/376208a0), on the twentieth of July, a month after Shell had already surrendered, which tells you what the science was worth by then. Nobody in Germany refusing to fill up their car was doing a dilution sum. They were answering the question on the placard, and to that question the arithmetic was not an answer. It sounded like an excuse. ## The part everyone forgets Here is the part almost everyone who tells this story forgets. That September, Peter Melchett, who ran Greenpeace in Britain, wrote to Shell’s UK chief executive and [apologised](https://www.ogj.com/home/article/17217152/greenpeace-we-erred-in-brent-spar-controversy). Greenpeace’s own sampling had been wrong, and the oil figure it had cited was a gross overestimate. A later audit by Det Norske Veritas, the Norwegian assurance body, commissioned by Shell, found the same. And the Spar itself was eventually towed to Norway and recycled, its hull cut into rings and used as the foundation for [a ferry quay near Stavanger](https://www.edie.net/brent-spar-dismantling-begins/). Onshore. The very outcome the campaign had demanded, reached years later, by which time nobody was watching. So Shell was right. Right on the arithmetic, right that the deep sea was the least bad option, and vindicated in writing by the very people who had beaten it. And it lost anyway. Utterly. Worse than that: within three years the [dumping of disused offshore installations at sea was outlawed](https://www.ospar.org/documents?v=57705) across the whole North-East Atlantic, bar a few narrow exceptions, and Brent Spar was the reason. The precedent Greenpeace was actually fighting for, it won. Greenpeace was wrong about the facts and right about the argument. Shell was the other way round. That is the whole of Brent Spar in two sentences. ## The wrong lesson The industry took a lesson from all this, and it took the wrong one. What everyone carried away was: never fight Greenpeace. Do not pick a public battle with a campaign group, because they will win it whatever the facts say. For thirty years that has been the reflex, and it has produced a generation of companies that fold at the first sight of a placard. But that was never the lesson. Shell did not lose because it fought. It lost because it fought on the wrong ground. It answered the question it could win, how much oil, how deep, how harmful, when the question actually being asked was another one entirely: whether a company got to treat the sea as its own back yard. The real lesson of Brent Spar is not that facts do not matter. It is that you have to know which argument you are in before you decide which facts to bring to it. ## The same mistake, thirty years on Which brings me back to those advertising rulings. The regulator keeps telling Shell that its adverts are within the rules, and it keeps making no difference, because the public is not arguing about the wording of an advert. It is having the Brent Spar argument again, in a new decade, about whose side an oil company is on. Win the ruling, lose the room. And when the industry does still fight, it fights where it thinks it can win. Shell [took Greenpeace to court](https://maritime-executive.com/article/shell-sues-greenpeace-for-2-1m-and-seeks-injunction-after-at-sea-boarding) over a boarding of one of its vessels in 2023. The case ended two years later in a settlement: no admission of liability, and Greenpeace paying three hundred thousand pounds not to Shell but to the lifeboats. Shell was within its rights to bring it. It still came out looking like a company that takes campaigners to court, which is precisely the room it could not afford to lose. The table again. You start to see the shape everywhere once you know to look for it. A technology company answers a town frightened of a data centre by setting out, to the litre, how much water the site will draw, and cannot understand why the number persuades nobody, because the town was not asking an arithmetic question. It was asking who the place is for. A carbon-neutral claim is defended on the accounting while a court in California lets [a case against Delta](https://www.climatecasechart.com/collections/berrin-v-delta-air-lines-inc-_0837b4) proceed anyway, because the dispute was never about the accounting. In each case the company is winning the argument on the table and losing the one in the room, and reaching, as it loses, for more of the very thing that is not working. Another figure. Another clarification. Another footnote. None of this is a brief for Shell. The same year as the Brent Spar, a far graver story was unfolding around the company in Nigeria, where the writer [Ken Saro-Wiwa was hanged that November](https://www.britannica.com/biography/Ken-Saro-Wiwa). That is a different essay, and a heavier one. This one is only about a single argument, and how a company that was right still contrived to lose it. ## Being right is not a plan So by all means be right. Get the numbers straight, commission the audit, know your own case better than the other side knows theirs. But do it understanding that being right is the price of entry and not a plan, and that the company which wins every fact and loses the room has, once the noise dies down, simply lost. Shell proved it was right about the Brent Spar. It took a written apology, an audit and a quay near Stavanger to do it, by which time the argument was long over and had gone the other way. Thirty years on, a regulator keeps confirming that the advertising is within the rules, and the room keeps not caring. Somewhere in the gap between those two things sits the whole difference between winning and being believed. Most boards still have not learned it. They are too busy being right. The Sutton Memo is a newsletter on reputation and PR-agency deals, from someone who has spent three decades on both sides: crisis adviser to boards, and M&A adviser to agency founders and buyers. Some weeks it is the crisis. Some weeks it is the deal. Tim Sutton advises the owners of communications agencies on their sale, and boards in a crisis. [timsuttonpr.com](https://timsuttonpr.com) --- # What a buyer pays for _PepsiCo moved its media after more than twenty years without running a pitch. Here is what that does to the price of a PR agency._ Date: 2026-09-10 URL: https://timosutton.substack.com/p/what-a-buyer-pays-for Earlier this month PepsiCo moved its media account. Not a small thing. [More than two hundred markets](https://www.mediaweek.com.au/publicis-takes-pepsico-s-us-1-7bn-media-account-from-omnicom), and a company whose 2025 accounts put its [advertising expense at $3.4 billion](https://www.sec.gov/Archives/edgar/data/77476/000007747626000007/pep-20251227.htm). What makes it worth your attention is not who won. It is how. There was no pitch. PepsiCo did not ask a row of agencies to walk into a room and win the business over ninety minutes. It ran what it called a capabilities review, looked at what the contenders already were, and appointed. Omnicom, which had held the account in the United States and Britain for more than twenty years, [lost it, and its shares fell about five per cent within hours](https://www.mediapost.com/publications/article/417647/omnicom-shares-fall-after-loss-of-pepsico-account.html). For most of my career the pitch was the arena. It was where an agency proved, in a room, that it deserved the business, and most firms in this industry still model themselves on being brilliant in that room. If the largest accounts now move instead on a review of what you already are, then the thing being bought is not the ability to win the room. It is what you were doing before anyone asked. And unlike a pitch, that cannot be written the week before. I want to follow that thought somewhere most of the coverage did not, because the coverage was about Omnicom and Publicis and the share price, and the more interesting question is what it means for the far smaller agency you may happen to own. ## The arena is closing at the top Start with what a capabilities review actually is, because the name is doing quiet work. A pitch asks a question about the future: what would you do for us, how clever can you be in the next hour, dazzle us. A capabilities review asks a question about the past: what are you already, what have you already built, what does your record say you can be trusted to run. The first rewards performance. The second rewards the standing record. A capabilities review is not nothing. There is still a room, and agencies still present in it. But what they present is what they have, not what they would do. PepsiCo, after two decades of the old way, chose to judge the record. One account is not a trend, and I am not going to pretend it is. Below the mega-clients the pitch is alive and will stay alive, and plenty of good business still moves the old way. But the direction of travel at the top is not in much doubt, and the top is where the tone is set. When an advertiser this size decides the performance was never the point, everyone downstream eventually hears it. There is a tell in the same announcement. Publicis [stepped back from the Coca-Cola process](https://www.adweek.com/agencies/publicis-lands-pepsicos-global-media-business-withdraws-from-coke-pitch/), because you cannot hold both, which tells you which account it valued. This was a portfolio decision, taken on the standing book of business, on a balance sheet rather than in a pitch room. ## Now put a price on it I [wrote a couple of weeks ago](https://timosutton.substack.com/p/who-is-buying-agencies-now-and-why) about who is buying agencies now. This is about what they pay for. Here is where it stops being someone else’s industry news and starts being about your number. I have spent thirty years on both sides of this. I ran the profit and loss on a large agency business for the best part of seventeen of them, I have sat on the winning and the losing side of more pitches than I can count, and I have bought agencies. And when you buy one, you do a private version of exactly what PepsiCo just did in public. You run a capabilities review. You look at what the firm already is. ## The review no client runs You may say media is the easy case. Media is scale and technology and buying clout, which can be read off a page, and public relations is judgement and relationships, which cannot, so the pitch will survive in our trade long after it has gone from theirs. Perhaps, though that page is getting easier to read in our trade than it was, and that is another essay. But notice that it does not matter. Whether or not a client ever runs a capabilities review on you, the one review that sets your number already happens, and no client runs it. A buyer of an agency has never paid for the pitch. They sit through the management presentation politely, and then they price the book. ## The book, not the pipeline What a founder is often proudest of, a buyer discounts almost to nothing. The pitch pipeline. The near-wins, the final two, the account we are sure to land next quarter. A buyer pays little for that, and is right not to, for two reasons. It is speculative, and it is portable: the pipeline tends to walk out of the door inside the heads of the three people who generate it, and those three people are the ones most likely to leave once they have been paid. What a buyer pays for is the opposite of the pitch. It is the revenue that is already retained, already contracted, already recurring. It is the client on year seven who renews without putting the account out to review, because that client is the proof that you do not have to win the room every time to keep the business. This is why the earn-out exists, and why it will not go away however much founders dislike it. The earn-out is the buyer saying, in the politest possible legal language, that they will believe your pipeline when it turns into retained revenue and not one day before. It is a machine for pricing the same thing PepsiCo was pricing: not what you might win, but what you can be trusted to keep. Agency valuations turn on the quality of the revenue base, on client tenure and concentration and how much of the book renews without a fight, far more than on the win rate a founder likes to quote. ## The pitch machine is a depreciating asset Follow that to its uncomfortable conclusion. If the whole story of your agency is that you punch above your weight in pitches, you are selling a skill that the largest clients in the business are starting to route around, and one that a buyer of your agency was never going to pay full price for in the first place. A brilliant pitch team is real, and I would not be without one. But as something to sell it is fragile. It depends on specific people, it produces revenue that is won and therefore can be lost the same way, and it is precisely the capability that the market at the top is quietly deciding it can do without. ## The record compounds slowly The standing record is the reverse in every respect. It is the seven-year client and the contract with a notice period and the retained scope and the reputation in a sector that gets you invited rather than auditioned. It is durable, it is not lodged in one person’s charm, and it compounds. It is also, and this is the catch, slow. You cannot manufacture it in the quarter before you sell. If you are two years from a conversation with a buyer, most of what will set your number is already fixed, and the shiny pitch you win next month barely moves it. The record was being written, or not written, for years before anyone made you an offer. ## Yes, but Let me put the other side, because it is a fair one. Pitching is not dead, most agencies are nowhere near the size where a client would run a capabilities review on them, and for the mid-market firm a strong new-business engine is a genuine part of the value, sometimes the largest part. And for a young agency the pitch is not optional. It is how the record gets written in the first place, because you cannot retain a client you never won. All true. To a point. But you are not pricing your agency for the market as it is this morning. You are pricing it for the buyer you sit across from in three years, or five, and I would not bet a sale price on the direction of travel reversing between now and then. The safe assumption is that what is rewarded at the top today is what is rewarded further down the market by the time you come to sell. ## The version that is for sale So here is the whole of it. What a buyer pays for is the version of your agency that exists on the morning when nobody is pitching for anything. The retained clients still retained. The contracts still running. The record still standing. PepsiCo has just shown the entire industry, in public and at the very top, that this is the version that gets chosen, and that the performance was always the smaller part of the story. The pitch was the arena. The record is the asset. If you are going to sell this business one day, spend rather less of your energy proving you can win the room, and rather more on what happens in the years after you have won it, which is the one part of your agency that cannot be written the week before. The Sutton Memo is a newsletter on reputation and PR-agency deals, from someone who has spent three decades on both sides: crisis adviser to boards, and M&A adviser to agency founders and buyers. Some weeks it is the crisis. Some weeks it is the deal. Tim Sutton advises the owners of communications agencies on their sale, and boards in a crisis. [timsuttonpr.com](https://timsuttonpr.com) --- # AI is not coming for PR. It is coming for the agency. _The machines eat the work that paid the bills. The judgment they cannot touch is the part we always undercharged for._ Date: 2026-09-11 URL: https://timosutton.substack.com/p/ai-is-not-coming-for-pr-it-is-coming Everyone in my industry has now read the same obituary. Artificial intelligence, we are told, will hollow out public relations: the machines will write the release, build the media list, draft the crisis statement, and the agencies will shrink to nothing. It is a tidy story, and I think it is wrong in the way that matters. It is wrong because it mistakes what an agency sells. An agency does not really sell press releases. It sells them as a pretext. What it actually sells, and the way it actually makes its money, are two different things, and AI does very different work to each. One it is about to make nearly free. The other it cannot touch. Pull the two apart and the whole picture inverts. AI is not an extinction event for public relations. It is an extinction event for the agency business model, and close to the best thing that has happened to the individual adviser in a generation. I have argued in [an earlier essay](https://timosutton.substack.com/p/is-pr-still-worth-it) that AI is taking the bottom of the ladder and leaving the top more room. What I did not say then was what that does to the business the ladder was holding up. Let me try to show you why, because the argument only works if you know where the money in an agency actually comes from. I spent the better part of two decades finding out. ## What an agency actually sells Strip the romance away and a communications agency is a leverage business, in the precise sense the word has in professional services. You hire bright people young and relatively cheaply, you bill their time out at a multiple of what they cost, and the profit is the spread, widened across a broad base of juniors sitting under a narrow point of seniors. The pyramid is the machine. Utilisation, leverage and rate are the three dials, and a well-run shop keeps all three honest. Most of the industry dresses this up as a monthly retainer, but a retainer is a timesheet with the numbers taken off: a bundle of hours priced in advance and, in most agencies, quietly over-delivered every month. For most of my career a twenty per cent margin was not a target but a discipline, the number that told you the pyramid was the right shape. I ran that model across nineteen offices and eighteen hundred people, and took one regional practice from $75 million in revenue to more than double that. I am not nostalgic about it. I am telling you how the sausage was made. Here is the part we were always too polite to say out loud. The client thought they were paying for the counsel, the judgment, the grey head in the room when it mattered. They were, a bit. But the invoice was built on the hours underneath: the first drafts, the media lists, the monitoring reports, the competitor scans, the slide decks, the background notes that three analysts produced over a weekend. That was the billable weight. The senior’s hours were a rounding error on the invoice, and much of the counsel that mattered was given in the unbilled hour anyway: the late call, the weekend, the conversation in the car. The work paid for the wisdom. We sold the one by the hour and gave the other away on top. ## The layer AI eats is the layer that paid Now look at what the machines are genuinely good at, today, not in some promised future. They are good at exactly that underneath layer. The first draft, the media list, the monitoring sweep, the research memo, the competitor read, the holding statement worked up from a template: this is the work AI now does quickly, cheaply and well enough, and does better every quarter. The machine does not need to be brilliant. Sufficient will do, and sufficient now costs almost nothing. We have heard that the machines are coming before, and the agency survived the web, social media and a decade of monitoring software. Those waves automated the edges of the business, how the work was distributed and measured. This one automates the middle, the production of the work itself, which is where the billable hours actually lived. You cannot bill forty hours for a piece of work the client watched a model produce in four. So the hours at the base of the pyramid, the hours that carried the spread, are the hours that fall away. The base narrows. And a pyramid that loses its base does not become a smaller pyramid. It becomes a spike, which is a different shape with very different economics. The margin that lived in the leverage goes with it. ## You cannot efficiency your way out The comfortable answer, the one I hear in every agency boardroom, is that we will simply use the tools ourselves, do the work faster, and keep the margin. It is a lovely thought and it does not survive contact with the client. The trouble is that the client can buy the same model you can. Once a capability is cheap and universal, its price collapses to the cost of the tool plus a thin markup for the bother, and the value migrates to whatever is still scarce. This is the oldest lesson in any leverage business: every efficiency you find is eventually competed away to the buyer. Being first to make your own product free is not a strategy, however energetically you put it on a conference slide. There is a better version of the objection, so let me give it its due. Perhaps the juniors do not disappear; they move up, and become editors and supervisors of the machine, checking its work and steering it. Some will, and that is real. But supervising a model is a thinner layer than producing the work by hand, it needs far fewer people, and it bills at a fraction of what the old production army did. The pyramid survives in outline and stops making the money. Either way, the spread is gone. ## The gift So far this reads as an obituary after all, only for the agency rather than the profession. But look at what is left standing once the production floor clears. What remains is the thing that was always scarce and never really commoditised: the judgment of what to do, said to the right person, in the hour that decides everything. I have [written here before](https://timosutton.substack.com/p/ai-has-already-read-your-crisis-statement) that information is the one thing a crisis is never short of, and that the machines widen the flow of it while narrowing the context you were actually missing. That call, made under uncertainty and legal exposure, with a chief executive looking at you and a reputation in the balance, is the one task in this business that does not delegate. Not to a junior, and not to a model. Nor, for the most part, to the client’s own people, however good. The person inside the building is inside the building: subject to the same fear, the same politics and the same reporting line as everyone else at the table. What a board buys from outside is the person who has seen this week before, many times, in other rooms, and owes nothing to anyone in this one. No amount of tooling manufactures that. That is the work clients always said they were buying. The difference is that they will now have to pay for it directly, on its own, without a month of billable production to hide it inside. And the person best placed to sell it is not the large agency with a floor of people to keep busy and a bench to feed. It is the senior adviser who carries no pyramid at all, whose cost of producing the supporting work has just fallen to very nearly zero, and who was only ever being bought for the judgment in the first place. The counsellor loses nothing here. What goes is the overhead that used to sit between the counsellor and the client, and with it the excuse for not paying the counsellor properly. ## What this does to the deal If you own an agency, this changes what you are selling, and I say that as someone who now spends a good deal of his time advising founders on exactly that sale. For years a buyer acquired a staffed delivery engine and paid a multiple on its billable leverage. As that leverage thins, the buyer is increasingly paying for something else: the relationships, the reputation, the judgment, the brand. People and trust, in other words, rather than a machine. That makes [the founder-dependency problem](https://timosutton.substack.com/p/what-does-a-buyer-actually-own-when) worse, not better, which is not what most owners expect to hear. Take away the pyramid and the value of the business concentrates even harder into a few senior heads, and those heads are the one asset a buyer cannot lock in the safe overnight. The remedy is seniority spread wide, the relationships and the judgment held across several names rather than one, and it has nothing to do with adding heads. A firm of a dozen, five of whom clients would follow anywhere, is a business. A firm of sixty carried by one founder, or by a floor of juniors the machine has just made surplus, is a personal practice with a payroll. Headcount, long treated as a sign of scale and a support to the valuation, starts to look like a liability: a payroll that no longer bills what it costs. The numbers that will move a price are going to be revenue per head and the ratio of senior to junior, not the total count on the door. The agencies that sell well over the next five years will be smaller, more senior and higher-margin than the ones that sold well over the last five. Bigger will not mean more valuable. In places it will mean less. None of this is a reason to despair, unless what you were selling was the pyramid itself. AI was never going to replace the person whose job is to be trusted in the worst week of a company’s life. It is simply going to stop the rest of us pretending that was not what we were charging for all along. The Sutton Memo is a newsletter on reputation and PR-agency deals, from someone who has spent three decades on both sides: crisis adviser to boards, and M&A adviser to agency founders and buyers. Some weeks it is the crisis. Some weeks it is the deal. Tim Sutton advises the owners of communications agencies on their sale, and boards in a crisis. [timsuttonpr.com](https://timsuttonpr.com) --- # Reputation needs more than a sentiment score _Why boards need to look beyond social media sentiment to understand whose trust is changing, and what that means for the business._ Date: 2026-09-13 URL: https://timosutton.substack.com/p/reputation-needs-more-than-a-sentiment After Donald Trump visited a Pennsylvania McDonald’s franchise on 20 October 2024, the company [reminded employees that it did not endorse candidates for elected office](https://apnews.com/article/mcdonalds-trump-campaign-harris-fries-56a5773528e212df058f85ec0f264578). Its formulation was memorable: “We are not red or blue – we are golden.” The line drew on a familiar brand to address a question of political neutrality. It was a statement about the company’s position, rather than a promotion of its products. Brand equity and corporate reputation are often discussed as though they amount to much the same thing. Brand equity is the value a brand adds to customer choice and willingness to pay. Reputation concerns how the people and institutions on whom a business depends judge its conduct, competence and credibility. A strong brand can help a company defend its reputation. It cannot answer every question about whether the business deserves people’s trust. In Bloomberg Media’s [2025 Corporate Reputation study](https://assets.bbhub.io/media/sites/2/2025/09/Corporate-Reputation_Bloomberg-Industry-Accelerator-2025.pdf), a survey of 1,250 senior business decision-makers in the US, UK, Hong Kong and Singapore, 79 per cent described reputation and brand equity as somewhat or entirely distinct within their organisations. The challenge is to carry that distinction through to what companies measure and put in front of their boards. ## What the dashboard leaves out In the same report, social media monitoring was the most commonly selected reputation metric: 36 per cent, compared with 25 per cent in 2024. Social media monitoring covers more than sentiment scoring, and the survey does not establish how much weight companies give these measures. My concern is what happens when a sentiment chart stands in for an assessment of the company’s reputation. The appeal is obvious. Marketing teams are used to regular reporting, and social listening produces a steady supply of material: changes in sentiment, spikes in attention, narratives gaining ground. It can alert a company to an emerging issue and expose concerns that management needs to hear. Online anger can have serious commercial and political consequences. The difficulty comes when that information is presented as a measure of reputation more broadly. The people posting are not a representative sample of everyone whose trust matters. A large volume of criticism may come from people with little connection to the business, while a concern shared by a handful of important customers may barely register online. A board looking at a sharp fall in sentiment needs help interpreting it. Are existing customers losing confidence? Is a regulator questioning the company’s account of events? Are employees worried enough to leave? Without that context, the chart leaves the board to guess at the significance of the movement. ## Start with the relationships Reputational damage can build for years or arrive in an afternoon. To understand its likely consequences, a company needs to know where confidence is weakening and how that might affect decisions. Much of the useful evidence will come from ordinary business conversations and records: Customers and commercial partners: Whether they stay when the headlines turn hostile, renew contracts on the usual terms, delay decisions or ask for new safeguards. Employees: Whether valued people are staying and candidates are accepting offers, alongside what staff say privately about the leadership. Regulators: Whether they remain confident in the company’s information and controls, or seek additional assurance and scrutiny. Lenders and investors: Whether they regard an issue as contained or see a failure of judgement that changes their willingness to commit capital. None of these is a clean measure of reputation. Contracts are lost on price, people leave for personal reasons and regulators have statutory duties. Interpreting a change requires a baseline and some investigation. If a partner delays a renewal, someone needs to ask why rather than assign it a reputational score. That work should be happening before a contract is lost. Regular stakeholder research, supported by direct conversations, gives a business a chance to pick up concerns while there is still time to respond. Social listening can help guide those enquiries; it cannot do the whole job. ## Give the board something it can use This need not become an argument about departmental boundaries. What matters is that the reporting answers the questions the board has to act on. A useful report would bring the online picture together with stakeholder research and evidence from the business. It would explain whose confidence appears to be changing, what supports that assessment and what remains uncertain. Management should then be able to recommend a response, whether that means explaining a decision more clearly or changing the conduct that caused the concern. A sentiment chart can belong in that report. It should not be expected to stand in for an assessment of the company’s reputation. By all means watch the timeline. Just do not mistake it for the business. The Sutton Memo is a newsletter on reputation and PR-agency deals, from someone who has spent three decades on both sides: crisis adviser to boards, and M&A adviser to agency founders and buyers. Some weeks it is the crisis. Some weeks it is the deal. Tim Sutton advises the owners of communications agencies on their sale, and boards in a crisis.[ timsuttonpr.com](https://timsuttonpr.com) --- # There is an invoice _The price of saying no to a client is real. So is the price of never saying it._ Date: 2026-09-14 URL: https://timosutton.substack.com/p/there-is-an-invoice In August, KKR [agreed to pay $250 million](https://www.justice.gov/opa/pr/kkr-agrees-pay-record-250m-penalty-serial-violations-federal-premerger-review-law) to settle a Justice Department case alleging that it had repeatedly failed to comply with the US pre-merger notification rules, across at least sixteen transactions. It is the largest penalty ever levied under those rules, by some distance. The detail that caught my attention was who, according to KKR, would end up paying it. [Reuters reported](https://finance.yahoo.com/markets/stocks/articles/kkr-settles-us-antitrust-case-015103051.html) that the firm expected its outside law firms to reimburse the penalty in full, leaving KKR, its funds and its investors with no financial impact. Two things need saying before anything else. KKR disputes the government’s characterisation of the case, says it acted in good faith throughout, and admits no liability in the settlement. And nobody outside the room knows what advice passed between the firm and its lawyers, or why the reimbursement was agreed. This is not a story about lawyers telling a valuable client what it wanted to hear, and I am not going to pretend it is. It is a story about power. Whatever the merits, when a client can tell its investors that its advisers will carry a nine-figure penalty in full, you have learned something about who needs whom in that relationship. A client that expects its advisers to absorb a quarter of a billion dollars for it is not a client those advisers will find easy to say no to. In my own trade we spend a great deal of time discussing the value of independent counsel. We spend rather less on what allows an adviser to remain independent when the client matters that much. ## The part I left out I contributed to that omission myself. In [my last essay here](https://timosutton.substack.com/p/ai-is-not-coming-for-pr-it-is-coming) I described the outside adviser as someone who had seen the crisis before and “owes nothing to anyone in this one”. That was too easy. There is an invoice, for a start. Independence, as the word is used in my profession, usually means one thing: the adviser does not report to the chief executive. That is true, and it is the least important of the three things independence has to mean. The adviser may be free of the chief executive’s reporting line and still depend on the chief executive’s goodwill for a substantial part of next year’s income. A small senior practice has fewer salaries to pay than a large agency, and losing one client hurts it a great deal more. Neither the size of the firm nor the word independent above the door settles the question. The balance sheet does, and so does one other thing I will come to: who, inside the client, is allowed to hear the advice. ## The account everyone watches There is no need to imagine a dramatic instruction to conceal something. The ordinary pressure is enough. Suppose an agency is advising on a launch. The date has been announced, the chief executive is committed and the client team has spent months preparing. The adviser thinks the company cannot yet substantiate a claim at the centre of the campaign. The recommendation should be to postpone, or to change what is being promised. That is an awkward conversation at any time. Add a renewal in six weeks and a team whose jobs depend on the account, and it acquires another audience. Before the recommendation reaches the client, the adviser has to explain it to colleagues who will bear some of the cost if the relationship goes wrong. Nobody has to order the advice changed. A firm can gradually teach its people which recommendations need another internal discussion, which should wait until the client is in a better mood, and which are best delivered so gently that the recipient can miss them altogether. Tact is part of the job. An adviser who cannot make an unwelcome recommendation usable is not much use. But there is a point at which softening the delivery changes the advice. A client who has been told there are some issues to work through has not necessarily been told that you think the launch should stop. ## Who gets to hear the advice? In corporate communications, the person who hires the adviser is acting for a company. Their interests will usually coincide with the company’s. Sometimes they will not, particularly when the matter under discussion is a decision they have already made. That is when access matters. If the executive whose decision needs challenging controls every conversation with the board, the adviser’s independence has a practical ceiling, however impressive the scope written into the engagement. The chief executive still decides who hears what. A board buying independent counsel should settle this before there is a dispute: who receives a serious adverse recommendation, whether the adviser can speak to the chair without first asking permission from the person being challenged, and which concerns must be escalated, and by what route. None of this is a licence for an adviser to go behind management’s back whenever they lose an argument; the route has to be understood by everyone at the outset. Discovering during a crisis that the board thought it was buying an independent assessment, while management thought it was buying help delivering an agreed position, is an expensive misunderstanding. There is a responsibility on the receiving side too. A chief executive can ask for candour in the appointment meeting and then make it plain that disagreement will cost the adviser access. The second message is the one that governs the relationship. ## Saying no is not proof of good judgment Advisers can be wrong. They can overestimate a reputational risk, misunderstand the business, or recommend caution because caution is safer for them. The cost of a delay falls on the client. It is quite easy to be uncompromising with somebody else’s money. So the unwelcome recommendation has to earn its place. Explain the evidence, what is uncertain, and what is likely to happen if the company proceeds. Offer a workable alternative where there is one. Be clear about what would change your view. Then let the people responsible make the decision. A client rejecting advice is not, by itself, a failure of the relationship, and not every disagreement requires a resignation. There is considerable value in an adviser who stays, accepts a lawful decision they argued against and helps the client manage the consequences. Refusing to put your name to a claim you cannot support is a different matter, and the distinction needs drawing carefully. Otherwise independence becomes an excuse for inflexibility, and the client has every reason to buy its advice elsewhere. ## What an owner can afford The difficult part for an agency owner is making this work when the numbers are tight. A speech about courage will not reassure an account director who believes that losing the client will cost colleagues their jobs. Start with the largest accounts. Work out what their loss would do to cash, staffing and profit over the notice period. Client concentration belongs in this discussion as much as it belongs in a buyer’s due diligence, and for the same reason: if one account decides whether the firm can meet its obligations, everyone handling it is working under that constraint, whether the owner acknowledges it or not. Then look at how people are rewarded. If retaining and growing an account are the only results that count, do not be surprised when staff hesitate to recommend anything that threatens either. An owner who agrees that a difficult recommendation is necessary has to own its commercial consequences. The account team cannot be told to speak plainly on Monday and marked down for the lost revenue on Friday. None of this contradicts the case I made for small senior firms. The firm with no pyramid to feed really is better placed to sell judgment. It is also more exposed to the loss of any one client, and the two facts sit side by side. What separates a practice that can afford to be independent from one that merely says so is whether the owner has priced the loss of the largest account before it happens, rather than discovering the price in the meeting where it is threatened. ## Where the bill comes due Long relationships can carry very candid advice. A client who has repeatedly benefited from an adviser’s judgment has a reason to listen when the next recommendation is unwelcome. There is nothing compromising about wanting to keep the business. The trouble begins when keeping it becomes a condition of what you are prepared to say. And that cost is not paid in the meeting. It is paid later, in the only currency an adviser has, which is being believed when it matters. I still think clients will pay for experienced, independent judgment. But I would put a harder question to any owner proposing to build a business around it, including the small senior firms I argued for. What happens to your firm when that judgment puts your largest account at risk? Your answer will tell the people working for you how independent they can afford to be. The Sutton Memo is a newsletter on reputation and PR-agency deals, from someone who has spent three decades on both sides: crisis adviser to boards, and M&A adviser to agency founders and buyers. Some weeks it is the crisis. Some weeks it is the deal. Tim Sutton advises the owners of communications agencies on their sale, and boards in a crisis. [timsuttonpr.com](https://timsuttonpr.com) --- # Who was Shandwick? _Founders spend months on the price and an afternoon on the name. Here is how long the name survives, and what the day it goes tells you._ Date: 2026-09-15 URL: https://timosutton.substack.com/p/who-was-shandwick For the best part of seventeen years my business card carried a name whose origin I never thought to question. Weber I could place: Larry Weber, who built a Boston technology PR firm into one of the biggest in the world from 1987. Shandwick I could place too, as a firm: founded in London in 1974 by Peter Gummer, later Lord Chadlington. What I could not have told you was where the word itself came from, whether Shandwick was a man, a place or something invented over lunch. The answer, it turns out, is that Gummer took it from his father. Canon Selwyn Gummer, a clergyman, had supplemented a priest’s stipend by writing sermons and selling them to other vicars under the imprint of the Shandwick Publishing Company, the children assembling the copies at the kitchen table. Where his father found the word I still do not know, and it no longer matters. When Peter came to name his own firm he reached for his father’s, and he did it on purpose. He could have called it Peter Gummer and Associates. He chose not to, [he later explained](https://www.buzzsprout.com/97473/episodes/630325-lord-chadlington?t=205), because he did not want a firm in which every client expected to deal with him personally. He wanted an institution, not a man. Which is the whole of this essay, reached before the firm had taken on a single client. The founder who wanted an institution rather than a practice was the one who refused to put his own name on the door. And his is one of the very few names in what follows still on a door anywhere. The rest of this is about why, because if you own an agency you will sell one of these one day, and the name above your door will be treated very differently from the way you treat it. ## An afternoon on the name In a sale, the name comes up late and briefly. Months go on the price, on the shape of the earn-out, on which clients count as contracted and which as hope. The founder’s name, the one thing in the room that is not on the balance sheet and never was, is dealt with in an afternoon, usually as a statement of intent rather than a term with teeth. The buyer says, warmly and meaning it at the time, that of course the name will stay. And it does stay, for a while. Then one day it does not, and the founder, who by then has usually been paid and gone, reads about its disappearance in the trade press like everyone else. I have watched people who negotiated ferociously over every point of consideration grieve a name they gave away for nothing, because it had not occurred to them that it was theirs to keep or lose. So it is worth knowing, before you sit down, what actually happens to these names, and when, and why. The record is clearer than you would think. ## Who buys you decides Here is the pattern, and it is not the one most founders fear. Being acquired does not, by itself, kill your name. What kills it is who acquires you. Sell to a holding company, one of the great advertising conglomerates, and your name is remarkably safe, for a remarkably long time. A holdco is in the business of owning names, not replacing them; a portfolio of brands is the whole point of the structure. Burson-Marsteller sold to Young and Rubicam in 1979 and kept its name for another thirty-nine years. Hill and Knowlton came into what is now WPP in 1987 and carried its two founders’ names, both men long dead, until 2024. Cohn and Wolfe was bought by Young and Rubicam in 1984 and was still Cohn and Wolfe thirty-four years later. Porter Novelli has traded under the names of Jack Porter and Bill Novelli inside Omnicom and its predecessors since the 1980s. On this evidence a founder’s name, sold to a holding company, is about the most durable thing in the industry. ## When the buyer has its own name Sell instead to an operating firm, a business that already trades under a name of its own, and the clock runs very differently. That buyer does not want a portfolio. It wants you inside its brand, and your name is a redundancy it will retire as soon as the integration allows. Financial Dynamics, one of the strongest names in the City, was bought by FTI Consulting in 2006 and had been fully absorbed into the FTI brand within five years. Powerscourt sold to the TPG-backed Morrow Sodali in October 2023 and was rebranded Sodali and Co by the following July, gone in under a year. When Finsbury, Glover Park and Hering Schuppener were folded together in 2020, and Sard Verbinnen added soon after, all four names, two of them among the most valuable in corporate and financial PR on either side of the Atlantic, were gone inside about a year, replaced by the initials FGS. There are exceptions, and they are recent: Teneo bought Tulchan in 2023 and has so far kept the name. On this record I would not lean too hard on so far. The size of the cheque has little to do with it. What matters is whether the buyer has a name it would rather use than yours. ## Names die at the merger, not the sale Even inside a holding company, the reprieve is exactly that. The name lives until the holdco itself decides to consolidate, and then it dies at the merger, not at the original sale, often decades later, when the founder is long gone and cannot even mourn it properly. You can watch this happening in real time, this year. In November 2025 Omnicom completed its purchase of Interpublic, my old group, putting two of the largest holding companies under one roof, a consolidation I [wrote about here](https://timosutton.substack.com/p/who-is-buying-agencies-now-and-why) a few weeks ago. Within three months, in February 2026, the merged group [announced](https://theprpost.com/post/13540/omnicom-merges-golin-and-ketchum-porter-novelli-joins-fleishmanhillard) what it would do with the overlapping PR firms it now owned twice over. Golin and Ketchum would merge, though for the moment, the announcement was careful to say, each would retain its own brand identity. Porter Novelli, forty years a name inside Omnicom, would become “a dedicated brand within FleishmanHillard”. If you have watched a few of these, you know how to read it. Retaining the brand identities is the reassurance offered at step one. Cohn and Wolfe was a brand within someone else’s structure for thirty-four years, right up until the morning it was not. A name demoted to a brand within something is a name in a waiting room. ## The initials phase There is a tell in how these deaths are staged, too. The name rarely dies as a name. It first shrinks to its initials. Burson-Marsteller and Cohn and Wolfe became BCW. Manning Selvage and Lee became MSL. Finsbury and the others became FGS. Initials keep the continuity of the letterhead while removing the people, and they are almost always a stage rather than a destination. BCW lasted six years before it too went, absorbed into the single word Burson in 2024. A set of initials is a name in palliative care. ## Kept, or deleted Notice, though, that not every founder’s name is erased when it goes. Some survive as qualifiers. Weber Shandwick still carries Shandwick, a name that was never a person to begin with; Cohn and Wolfe lived on for years inside BCW. Others are simply deleted: Marsteller vanished in 2018, and Harris was dropped from Golin in 2014, so cleanly that no one who joined this century would know it had ever been there. Which of the two happens to your name is, again, not really about you. It is about how much reputation the buyer still needs to borrow from it at the moment of the merger. ## The name that outlived the man Now the strangest entry in the record, and the most instructive. In January 2020 Harold Burson died, at ninety-eight, the last of the generation that founded the modern industry and the B in a name that had already been reduced to initials. Four years later, in 2024, WPP took those initials, BCW, merged them with Hill and Knowlton, and relaunched the whole thing as Burson. The name of a man four years dead was brought back to the top of the door, not out of sentiment but because, of everything in the cupboard, it was the name with the most reputation still attached to it, and none of the newer names had accumulated anything like as much. A name, it turns out, is an asset entirely separable from the person who made it, and a buyer will retire it, keep it, or resurrect it purely on what it is worth to them that day. And Burson’s story is not finished, which is the part a founder should sit with. In February the Financial Times reported that WPP’s board regarded Burson as the easiest of its businesses to sell, because it sits apart from the rest; in April The Times [reported](https://www.bandt.com.au/report-wpp-explores-options-in-sale-of-pr-comms-firm-burson/) that Goldman Sachs had been appointed to explore options for it, under a new chief executive with a plan to simplify the group. WPP has declined to comment. Burson is around six thousand people and something like a tenth of WPP. No buyer has emerged as I write. But follow the rule of this essay to its conclusion and you can even guess at the name’s fate. If a financial buyer takes it, a private equity house with no agency brand of its own to impose, the way KKR took FGS, then Burson survives, because the buyer has no name it would rather use. Harold Burson’s surname may yet outlive its third owner, precisely because the people most likely to buy it have nothing to put in its place. The name is safest in exactly the hands founders tend to trust least. ## What a founder can actually do So what do you do with this, if the name on the door is yours? First, treat it as a term, not a courtesy. If it matters to you that the name survives, negotiate for it: a minimum period before any rebrand, a say in the timing, a right to take the name back if they retire it. An intention expressed warmly across the table in the afternoon is worth nothing on the morning the holdco consolidates. Almost no founder asks for this, because almost no founder believes, on the day of the sale, that the name is in any danger. The record says it is. Second, read the name as a signal, in both directions. A buyer who keeps your name is telling you something, and it is not that it is fond of you: it has not yet finished moving your clients and your reputation onto its own balance sheet, and it still needs your name to hold them there. The day it retires your name is the day it has decided it no longer needs it, which is to say the day [the transfer you were paid for](https://timosutton.substack.com/p/what-i-learned-buying-agencies) is finally complete. If you are still inside the business, the retirement of your own name is the most honest performance review you will ever get. And third, if you would genuinely rather not go through any of this, consider not putting your name on the door in the first place. Peter Gummer did exactly that in 1974, and a good deal of the City worked it out long ago. Many of the most durable names in British corporate and financial PR are places and inventions rather than people: Brunswick, Portland, Finsbury in its day. A name with no person in it has nothing to bruise when it goes, and its founder has nothing to grieve. It is no guarantee of long life. Finsbury and Powerscourt were place-names too, and both died anyway, one dissolved into initials, the other into its buyer’s brand within a year. A borrowed place-name does not save your firm from an operating buyer who wants it gone. It only saves you from taking it personally. Which leaves Shandwick. Peter Gummer, naming his firm in 1974 after the business that had printed his father’s sermons, was not betting on the name’s longevity. He was deciding, before he had a client, that the firm should never depend on him, and a name that does not depend on a person is one its founder can part with without grief. Half a century on the name is still there, half of a compound neither founder set out to build, priced and carried and not yet written off. It has outlasted most of the names in this essay not because it was a word rather than a person, but because at every sale and every merger some buyer still found it worth borrowing. That is the only reason any name survives. Gummer’s advantage was that he had separated himself from his before anyone could do it for him. Most founders make that separation, if they make it at all, on the afternoon it comes up, when the name they thought was theirs turns out to have been on loan to the firm all along, and the firm was always going to be sold. The Sutton Memo is a newsletter on reputation and PR-agency deals, from someone who has spent three decades on both sides: crisis adviser to boards, and M&A adviser to agency founders and buyers. Some weeks it is the crisis. Some weeks it is the deal. Tim Sutton advises the owners of communications agencies on their sale, and boards in a crisis. [timsuttonpr.com](https://timsuttonpr.com) --- # Who's buying agencies, and what it means if you own one _A plain read of the deals of 2026 so far, written for the people who own the firms._ Date: 2026-09-18 URL: https://timosutton.substack.com/p/whos-buying-agencies-and-what-it If you own an independent agency, the biggest financial question you will ever face is one you can discuss with almost nobody. Your team cannot hear it, your clients must not, and your peers are your competitors. So most founders learn how buyers think only when one is already across the table, which is the most expensive moment to learn it. I would like to make that a little easier. Who buys agencies has changed this year, and it changes what a business like yours is worth, and to whom. Whether a sale is three years off, ten, or not on your mind at all, that is worth knowing early, for a simple reason: most of what a buyer pays for can be built, but only by an owner who knows what it is while there is still time to build it. So this is the first of a note I mean to write regularly. Nothing in it is inside knowledge. Every deal is public, most of it announced by the buyers themselves, and the only thing I add is what I think it signals. To open, a look back over 2026 so far, because the shape of a whole year tells you more than any single week. The short version: two things have happened at once, and they pull in opposite directions. The big holding companies have largely stopped buying and started tidying. The real buying has come from somewhere else: private-equity-backed specialists, quietly assembling platforms one agency at a time. If you own a good independent, that second group is now far more likely to be the one that calls. ## The giants are consolidating, not shopping WPP spent the early part of the year rewiring itself, folding Ogilvy, VML, AKQA and Burson under a single WPP Creative division in its [Elevate28](https://www.wpp.com/en/news/2026/02/strategy-update-and-2025-preliminary-results) plan, with around 500 million pounds of cost savings targeted by 2028. Omnicom merged two of its own media networks, Mediahub and Hearts and Science, into one operation of roughly 9 billion dollars in billings, now called [Hearts United](https://www.omc.com/newsroom/omnicom-media-officially-launches-hearts-united/). Stagwell brought Doner and Colle McVoy together as DonerColle Partners. Closer to home, The Mission Group merged krow into Bray Leino. Four of the largest names in the business, and none of those is the purchase of an outside company. They are housekeeping: fewer brands and lower cost, in a year when the holding-company model is under real pressure. The one large exception ran the other way, Publicis paying more than 500 million dollars for the sports-marketing agency [160over90.](https://www.publicisgroupe.com/en/news/press-releases/publicis-groupe-disrupts-sports-marketing-landscape-with-acquisition-of-160over90) A deal that size is a world away from most independents, but it says something about where the big groups still see growth, which is sport and entertainment. ## The real buyers have a fund behind them Underneath the giants, the acquisition volume this year came from specialist platforms with private equity behind them, running steady bolt-on strategies month after month. • Public affairs. The Public Policy Holding Company kept buying, taking WPI Strategy in London in March and The Advocacy Partners in Florida in August. That second deal is worth pausing on, because the numbers are public and rarely are: [20.4 million dollars up front and an earn-out worth up to a further 54.6 million](https://www.globenewswire.com/news-release/2026/08/03/3337357/0/en/PPHC-Acquires-Florida-Government-Relations-Firm.html), a maximum of 75 million in all. I will come back to what that split tells you. • Healthcare communications. Klick made its largest-ever acquisition, Oxford PharmaGenesis, its third in eighteen months. A healthcare-focused fund, Martis Capital, took a majority of Deerfield Group while leaving the founders with significant equity. Bridgepoint-backed Prescient bought Uptake in January. • Financial communications. ParkSouth-backed Infinite bought Dukas Linden in New York in May and Greentarget in London in June, doubling its London headcount in a single deal. • Branding. The private-equity firm WestBridge backed Koto’s first acquisition, Stereo Creative, the opening move in building a studio network. • And the corners you might not watch. A Mountaingate-backed roll-up called Podean bought a marketplace or social-commerce agency roughly every six to eight weeks, all year. None of these buyers is a household name. All of them are assembling something, and the agency they buy next will be the one that fits the platform, which is rarely the biggest on offer. ## New shapes, and buyers who do not mind where you are Plenty of this year’s deals were something other than a clean sale of the whole business. FUSE Create took a minority stake in the influencer agency Sway, which carries on running itself. Martis left Deerfield’s founders with real ownership. YKONE and Mirror Mirror merged rather than one swallowing the other. And the buyers increasingly do not care where you sit. Havas alone bought a sports-marketing agency in the Netherlands, the one that co-organises the Dutch Grand Prix, an experiential agency in Barcelona, and a majority of Acento in Iberian public affairs. Accenture Song paid up for [Whalar](https://newsroom.accenture.com/news/2026/accenture-to-acquire-leading-creator-and-social-agency-whalar-from-whalar-group), spanning the United States, the United Kingdom, Ireland, Germany and Spain, in one of the year's largest creator-economy deals.. Selling has stopped meaning one buyer writing one cheque and walking you out of the door. ## What I make of it The thread through all of it is who is buying, and why. If you own a good independent, the likeliest buyer is now a fund assembling a platform, and a fund buys differently from a rival. Last year’s profit matters less to it than two other things: whether your firm fits what it is building, and whether the business can stand on its own feet. That second test is harder than it sounds, because most founders have spent years, quite reasonably, building the firm around themselves. The encouraging part is that it can be worked on, and the work is mostly the same work as running the place well. It also changes what is worth watching. The headline price on someone else’s deal tells you almost nothing. The shape of the deal tells you nearly everything. The Advocacy Partners numbers are the tell: barely a quarter of the money on the day, and the rest riding on an earn-out over years. How much is paid up front, and how much depends on what happens next, is the real negotiation, and it is the one thing almost nobody puts in the announcement. It is also what decides, long after the headline, whether selling was the right thing to have done. I will keep writing this as the deals land. If it ever raises a question about your own firm, I am glad to be a sounding board. It costs nothing to ask. The Sutton Memo is a newsletter on reputation and PR-agency deals, from someone who has spent three decades on both sides: crisis adviser to boards, and M&A adviser to agency founders and buyers. Some weeks it is the crisis. Some weeks it is the deal. Tim Sutton advises the owners of communications agencies on their sale, and boards in a crisis. [timsuttonpr.com](https://timsuttonpr.com) --- # The best trade in PR _Sell your agency to a holding company. Wait. Buy it back. It has been done more often than the industry lets on, and almost nobody will say what they paid._ Date: 2026-09-19 URL: https://timosutton.substack.com/p/the-best-trade-in-pr At the end of August, in Berlin, two men bought their own company back. Benjamin Minack and Andreas Nickel had built an agency called Ressourcenmangel and, in 2009, sold it to the group that is now WPP’s Hirschen Group. For seventeen years it belonged to somebody else. Renamed Rysm earlier this year, it runs to some two hundred and fifty people across six offices. On the twenty-seventh of August the two founders [bought the whole of it back](https://www.marketingminute.co.uk/post/rysm-s-founders-buy-their-agency-back-from-wpp-s-hirschen-group), with an outside financing partner behind them and neither side saying what it cost. Minack’s account of why came to four words of German: Alleine geht es besser. It goes better alone. I want to take that sentence seriously, because it is the newest instance of a trade that has been available in this industry for twenty years and that almost nobody writes down. You sell your agency to one of the great marketing groups for a good price. You work out your time. Some years later, you buy it back. Described coldly, as a piece of dealing rather than a homecoming, it looks like one of the better trades in professional services. Whether it really is turns out to be unknowable, for a reason I will come to. The more interesting questions are why a holding company would ever let you do it, and why more founders are about to try. ## What a holding company actually buys Start with what changes hands in the first place, because it is not quite what the balance sheet records. When a group buys an agency it books goodwill, client contracts, a lease, some furniture. What it is paying for is the founder. In practice the group is renting the founder, and the earn-out is the term of the lease. The three years, or five, in which the founder cannot leave and cannot compete are the years the group needs to move the clients and the reputation off one person and onto its own books. [I have argued before](https://timosutton.substack.com/p/what-i-learned-buying-agencies) that an agency acquisition is, at bottom, the transfer of trust from a person to an institution. A buy-back is what it looks like when that transfer never quite completes. Because sometimes it does not. The clients stay loyal to the name on the door and the person who answered the phone. The institution never fully takes. The earn-out ends, and the group discovers it owns a business whose entire value can still walk out of the building on a Friday. At which point the founder occupies the one negotiating position nobody sets out to engineer: he is both the asset being sold and the only serious bidder for it. ## The quiet history of buying it back It has happened more often than you might think, and the roll-call is worth laying out, because you will know the firms. Matthew Freud sold a little over half of Freud Communications to Publicis Groupe in 2005. Six years on, in 2011, he [bought Publicis out](https://www.prnewswire.com/news-releases/publicis-groupe-sells-its-stake-in-freud-communications-119920334.html) and took the firm private again. Michael Kempner sold MWW to Interpublic in 2000, and [bought it back](https://www.prnewswire.com/news-releases/mww-group-completes-management-led-buy-out-from-interpublic-group-112926044.html) a decade later; ten years after that he renamed the firm MikeWorldWide, which is one way of settling for good the question of whose agency it is. Qorvis, the Washington public affairs shop, went into Publicis in 2014 and its management [led it back out](https://qorvis.com/pr-week-qorvis-completes-management-buyout-splits-from-publicis-groupe/) at the very end of 2022, Publicis declining, as these things go, to comment. APCO is a variation on the theme. Margery Kraus, its founding chief executive, was not the seller in 1991: APCO began as the affiliate of a law firm, Arnold and Porter, which sold control to Grey that year. But she made the buying half of the trade in full. On the thirteenth of September 2004 [Grey agreed to sell itself to WPP](https://www.wpp.com/en/news/2004/09/grey-global-group-agrees-to-join-wpp); a fortnight later APCO announced that its management had [bought the firm out](https://www.adweek.com/brand-marketing/apco-regains-independence-grey-75131/) through a merchant bank, on terms never disclosed. A buyout of that kind is not arranged in a fortnight, so it had evidently been in train for some time, but the effect was that when Grey passed to WPP, APCO did not go with it. And, closer to home, Frank. Graham Goodkind, who founded Frank PR with Andrew Bloch, sold it to the Australian group Photon, later Enero, in 2007. Over the following years the firm was bought back in pieces. The last of it [came home in 2021](https://www.prweek.co.uk/article/1709016/franks-leaders-buy-back-business-holding-group), when Goodkind and his managing director, Alex Grier, bought out Enero’s remaining three-quarters, with Bloch still a shareholder alongside them, for a figure that was, set beside the original deal, small change. What happened to the firm next is the part I like most, and I will come back to it. ## Why a group ever sells it back Notice what the public record does not show. It does not show a single one of these agencies being sold back because it was failing. What it shows, again and again, is a group in the middle of changing shape. Grey was selling itself. WPP, as I write, is simplifying itself under a new chief executive. A group refocuses on data, or on healthcare, or on whatever the analysts are rewarding that year, and a perfectly good agency that no longer fits the story becomes, in the language of the quarterly call, non-core. Non-core is not the same as unharmed. Set Frank’s last price beside its first and it is plain that the business was worth a great deal less to Enero at the end than at the beginning. But that is rather the point. An agency whose value lives in its founder is worth less in anyone else’s hands, and least of all in the hands of an owner whose attention has moved on. The tell is the silence. When these deals are announced, the group says as little as it decently can. Publicis wished Freud success in his future endeavours and offered no reason at all; asked about Qorvis, it declined to comment. The talking is left to the founder, who says something warm about independence and agility and nothing whatever about the price. Minack, with his four blunt words, is the exception. The reticence is understandable. If you have just bought back the thing you know better than anyone else alive, you do not stand in the street discussing what you paid. ## The price you are not allowed to see That reticence is more than good manners. It points at the strangest feature of this whole trade, which is that you cannot prove it is a good one. I went looking for a single case, among all of these, where both halves of the deal were public: what the founder was paid on the way in, and what he paid on the way back. There is essentially one, Frank, and even there the original 2007 price is reported two irreconcilable ways in the trade press, so the pair does not really close. On either figure the buy-back cost a fraction of the sale; what cannot be said is how large a fraction. Everywhere else, at least one number is missing. Freud’s sale price was reported; his buy-back price was not. For MWW, APCO and Qorvis I could find a public price for neither leg. The best trade in PR turns out to be the one trade whose profit is never printed. There is a certain logic to that. The one deal in which the buyer knows exactly what the asset is worth, down to the last client and the last account director, is the deal a founder does to buy back his own firm. It is the most honest valuation an agency ever receives, and it is the one the industry keeps in the dark. So the case for the trade cannot be made in cash. It has to be made the way the founders themselves make it, by pointing at a firm that has survived and grown under its own name again. That is evidence of a kind, and it flatters the trade. We only ever hear about the buy-backs where there was something left worth buying, and nobody issues a press release about the agency he decided not to rescue. ## Why this is about to happen more For twenty years the buy-back was a rare and slightly romantic move, the exception that proved how sticky these acquisitions usually are. I think it is about to become a good deal more common, and the deal in Berlin suggests why. Marketing Minute, reporting the buyout, has Minack framing independence as newly viable because AI now lets smaller shops compete with the networks on capability. That is one founder’s rationale, relayed by one trade report, and I would not hang an industry on it. But I think he is right, and the argument does not depend on him. Ask what a founder was really buying, in 1998 or 2005, when he sold to a network. Money, certainly. But also scale: reach into markets he could not open himself, a back office he did not have to build, the machinery that let a good small shop punch at the weight of a large one. For a long time you could only rent that machinery from a holding company. A good deal of it, the drafting, the research, the first pass at a plan, is now available to a fifteen-person firm for the price of a subscription. So the gap in capability between the independent and the network, which was much of the case for selling in the first place, is narrowing. (Rysm, I should say, is hardly a garage. It is a creative and digital agency with a communications arm rather than a pure PR shop, it has two hundred and fifty people, and alone turns out to include an outside backer. The principle holds all the same.) ## And the groups are shedding agencies On its own, that would only mean fewer founders selling. What turns it into more buy-backs is what is happening to the groups at the same time. Under pressure from the same technology, among other things, they are consolidating, simplifying and deciding what is core, and I have written in recent weeks about [how much of that is already under way](https://timosutton.substack.com/p/who-is-buying-agencies-now-and-why): Omnicom absorbing Interpublic, WPP reported to be exploring a sale of Burson. Every such tidying-up leaves agencies on the shelf marked non-core, and for each of them there is a founder or a management team doing Minack’s sum. The supply of agencies to buy back is rising at the same moment as the confidence to do it. The obvious objection is that the networks get the machines too, and so do the clients: the tools that level a small agency with a large one may equally level the client with both. I have no tidy answer to that. But it is an argument about whether agencies of any size keep their value, not about who is best placed to own one, and on that narrower question the drift seems to me to be towards the founder. ## What a founder can do about it First, negotiate the way back in on the way out. In my experience almost nobody does. In the year of the sale, when the cheque is large and the relationship is warm, the idea that you might one day want the firm back seems faintly ungrateful, so it goes unmentioned. Ask anyway. A right of first refusal if the group ever decides to sell or close the business costs the buyer little on the day and is worth a great deal on the morning the agency is declared non-core. It sits beside the point I made about [Shandwick](https://timosutton.substack.com/p/who-was-shandwick), that if you care whether the name survives you must make it a term and not a hope. The way home is the same: a term, or nothing. Second, know when the option is cheapest. It is cheapest when the earn-out has ended and the group is refocusing, because that is the moment you are worth more to yourself than to anyone else in the room. You can see that window coming a long way off. Third, read your non-compete as the price of the alternative. The alternative to buying the firm back is leaving and building another one, and the length and reach of the covenant you signed decides whether that alternative is real. Peter Gummer never bought Shandwick back after he sold it to Interpublic in 1998; he went and built Huntsworth instead. Larry Weber did not buy his firm back from the same group; he built what became Racepoint. Starting again is the other way home, and it recovers everything except the name, which is precisely the thing a buy-back does recover. Rysm’s founders took the opposite course: they kept the firm, name and all (a name, admittedly, not yet seven months old), and shed the owner. ## A line for the buyer I have sat on that side of the table too. If, in year six, the only credible bidder for the agency you acquired is the person you acquired it from, you have learned what you bought. You did not buy a business. You bought a lease on a founder, and the lease is up. ## Back to Frank Which returns me to Frank. Having finally got the whole of the firm back, Goodkind and Grier did not sit on it. In the summer of 2024 they moved control of Frank into an employee ownership trust. I do not know their reasons and will not guess at them, but the effect is neat. Any future buyer of Frank will be negotiating with a trust that holds the firm for the people who work in it, which is as close as company law gets to negotiating with the asset itself. It goes better alone, says the man in Berlin. I would amend him slightly. It goes better when the firm is owned by the people who are the firm, founder included, and the holding companies have spent a good deal of money, over a good many years, establishing the point. A postscript. The six buy-backs here are the ones I could verify. If you know of one I have missed, I would like to hear about it. The Sutton Memo is a newsletter on reputation and PR-agency deals, from someone who has spent three decades on both sides: crisis adviser to boards, and M&A adviser to agency founders and buyers. Some weeks it is the crisis. Some weeks it is the deal. Tim Sutton advises the owners of communications agencies on their sale, and boards in a crisis. [timsuttonpr.com](https://timsuttonpr.com) --- # Delighted to announce... _Nobody ever sells an agency. They join, they partner, they take the next step. The plain verb turns up eventually, and when it does it tells you what the soft ones were for._ Date: 2026-09-20 URL: https://timosutton.substack.com/p/delighted-to-announce There is a ritual to the sale of a public relations agency, and the strange thing about it is that it never mentions a sale. The founder posts a photograph, usually of himself, sometimes of a skyline. He is delighted, or thrilled, or, for reasons never explained, humbled. He has news to share. And then, in the announcement itself, the thing that has actually happened is described in every way but the one a dictionary would reach for. He has joined. He has partnered. He has begun his next chapter, taken his next step, found the right home for his people. Somewhere a larger company has agreed to pay him a sum of money for his firm, and he will spend the next few years earning the rest of it, but you will not hear that from him. You will hear it from the buyer. I have written a few of these myself, from the other side of the table, in the years I was buying agencies rather than advising the people who sell them. The drafting could sometimes be tortuous: trying to justify our strategic reasons for the acquisition and then trying to reconcile them with the very different perspective of the seller on the same transaction. And the biggest challenge of all was avoiding the usual clichés: the X enhances the Group’s ability to help clients... and the Y extends our offering across crucial capabilities including.... So I offer what follows as someone who has drafted the euphemism more times than I care to remember, not merely noticed it. ## Eleven announcements, and a clean sweep To be sure I was not simply remembering the deals that suited me, I did something anyone can do in an afternoon. I went through this year’s acquisitions of public relations and public affairs firms in one trade archive, [O’Dwyer’s](https://www.odwyerpr.com/story/category/101/acquisitions.html), in date order, and kept the deals in which the seller was quoted anywhere at all, in the trade report or in the release behind it. That gave me eleven. (I left out a twelfth because I know the people involved. It would not have changed the score.) I read each announcement twice. Once for the headline. Once for the words of the person selling. In all eleven the headline, the buyer’s own or the trade paper’s, used the plain word: acquires, or acquisition. In none of the eleven did the seller use sold, sale, acquired or bought, anywhere in the words attributed to them. Eleven to nil. As a sample it is small. As a scoreline it is emphatic. You are welcome to run the search yourself. ## What they say instead What the sellers said instead sorts into a small and repeating vocabulary. Five said they were joining, and two of those called it a next step in the same breath. One more called it a next step and left it there. One was partnering. One was being part of something larger. The remaining three kept to what I have come to think of as capability language: what the deal will let them do for clients, and how little will change for them, which quietly declines to say what kind of event it is. It is the seller’s cousin of the cliché I spent years trying to keep out of the buyer’s paragraph. They are worth hearing in their own words, because the drafting is good. Bob Schultz, selling The CauseWay Agency to 4media group, offered that [joining 4media group allows us to offer our clients expanded support](https://www.globenewswire.com/news-release/2026/06/02/3305127/0/en/4media-group-acquires-the-causeway-agency-expanding-government-nonprofit-and-public-affairs-expertise.html). Bruce Hennes of Hennes Communications [knew partnering with a larger agency would be of great benefit](https://www.prnewswire.com/news-releases/akcg--public-relations-counselors-acquires-cleveland-based-hennes-communications-302762847.html). Jon Rossi of Modern Advocacy found that [being part of Actum allows us to scale that work](https://www.odwyerpr.com/story/public/24690/2026-05-01/pr-firm-news-actum-acquires-nycs-modern-advocacy.html). John Anderson, whose firm went to FGS Global, fitted two of the soft verbs into a single sentence: [joining FGS is the perfect next step for RFA and for our clients](https://www.fgsglobal.com/insights/fgs-global-acquires-rich-feuer-anderson). Not one of them is lying. Each has sold a firm, and each has described the sale as a form of arrival. My favourite came from Stephen Shiver, co-founder of a Florida firm bought in August, who said that [reputation is earned over decades, not transactions](https://www.globenewswire.com/news-release/2026/08/03/3337357/0/en/PPHC-Acquires-Florida-Government-Relations-Firm.html). It is a good line, and I expect he means it. It is also about as close as any of the eleven sellers came to a plain commercial word, and the word is there only to be told that it does not matter. ## Two languages, one signature Here is the part that took me longest to see, for someone who has written these documents. The two vocabularies are not kept in separate rooms, the buyer’s in one and the seller’s in another. They sit in the same release, three paragraphs apart, under one date and one approval. Take Next Fifteen’s [announcement that it had bought M Booth](https://www.investegate.co.uk/announcement/bzw/next-15-group--nfg/acquisition/1761618), the New York agency, in 2009. Because Next Fifteen is a listed company, it is obliged to be plain with its shareholders, and it is: the initial consideration is $4 million, paid in cash at completion, with deferred consideration of up to a maximum of $13.25 million payable over the course of the next four years. That is the sound of a business changing hands. And then, further down the same page, the founder, Margi Booth, describes the same day in another language altogether. This, she says, is an important step in our evolution. The nouns of the acquisition and the language of the arrival, in one document, signed off together. Sometimes even the buyer will not say the flat word. When WPP [acquired The Glover Park Group](https://www.wpp.com/en/news/2011/11/wpp-acquires-the-glover-park-group-in) in 2011, its own release used the plainest possible verb in its headline, acquires the stock of, and then, a clause later, described the firm as joining the roster of WPP’s public relations and public affairs companies. The company writing the cheque reached for the very word the people cashing it prefer. Which tells you the euphemism is not the founder’s vanity. It is a joint production. ## The first work the deal does So what is it for? I argued in [an earlier piece](https://timosutton.substack.com/p/what-i-learned-buying-agencies) that what a group is really buying, when it buys an agency, is trust: the transfer of a relationship from a person to an institution. The soft verb is the first instrument of that transfer, and it does its work on the clients. Consider what the two words tell a client on the morning of the deal. Sold says the founder has been paid, the meter is running, and the person you hired has a new set of masters and a clock on the wall. Joined says nothing has moved. It is a holding statement, in the crisis sense, and like all good holding statements it is true, incomplete and designed to buy time: time in which no client reprices the relationship, no rival rings round your accounts to ask whether they are still in safe hands, and no account director dusts off her CV. I can say from the buyer’s chair that the buyer wants that stillness at least as much as the seller does, because the buyer has just paid for it and would prefer it did not walk out of the building. In my experience the most carefully negotiated sentence in an acquisition release is not the one about money. It is the one that says the founder will continue to lead. It is the sentence a buyer insists on and a seller is glad to give, and it is the truest thing in the document, for a while. ## The vanishing seller There is a related pattern worth noticing, and it concerns who gets quoted at all. When I went past the archive’s summaries to the full releases, I noticed something small and telling. The condensed version of a deal, the paragraph the trade press actually prints, keeps the buyer’s quote and drops the seller’s. Five of my eleven only yielded the seller’s words once I went to the release behind the report, and several other deals fell out of my count because the seller was not quoted anywhere I could find. The disappearance of the founder’s voice starts on day one, in the editing. ## The word arrives late The plain verb is not banished for ever. It is merely on a delay, and it is worth waiting for. Go back to Margi Booth. In 2009 the sale was an important step in our evolution. Years later, on LinkedIn, describing the same transaction to her own network, she began a sentence this way: “[when I sold M Booth to Next Fifteen over 10 years ago...](https://www.linkedin.com/posts/margibooth_when-i-sold-m-booth-to-next-fifteen-over-activity-7057104037149560833-dJlH)” Sold. The word that could not appear in the announcement turns up, unforced, once the announcement has long since done its job. I do not read that as a confession. It is simply the truth, told plainly by someone with no release to draft. She is not alone. Michael Kempner’s firm, MWW, was bought by Interpublic in 2000 and bought back a decade later, and [the buy-back release his own side wrote](https://www.prnewswire.com/news-releases/mww-group-completes-management-led-buy-out-from-interpublic-group-112926044.html) refers, without flinching, to the original sale. The original sale. I have not been able to find the announcement from 2000, which has gone the way of most press releases of that age, but I would be surprised if the word sale was in it. A decade on, the same firm, writing about the same deal, could call it what it was. The euphemism, it turns out, has a shelf life, and the shelf life is roughly the length of the reason it was needed. ## How long is that, honestly Here I have to be careful, because it would be easy, and wrong, to tell you that founders leave the instant the earn-out ends, as though the language ran on a timer. The record is messier than that, and I would rather show you the mess than a tidy line. Where the earn-out term is actually on the public record, the change at the top does tend to follow it. Red Consultancy’s founders stood down in 2003, when the trade press reported that their two-and-a-half-year earn-out had run its course. The deferred consideration on M Booth ran, on Next Fifteen’s own numbers, for four years from August 2009; the [report that its founder was moving to a chairman role](https://web.archive.org/web/20200813122546/https:/www.provokemedia.com/latest/article/m-booth-hires-new-ceo-as-founder-moves-to-chairman-role) is dated September 2013. At a Washington lobbying shop, the Federalist Group, some of the early partners, according to Roll Call, left once they had collected their portion of the sale through the earn-out. Three deals, three clocks, and in each the change came at about the time the last payment did. But set beside them the founders who make a nonsense of the pattern. Gershon Kekst sold his firm to Publicis in 2008 and never left: he was still there, as chairman emeritus, when he died in 2017. Al Golin was still working at the agency that bears his name, long since part of Interpublic, when he died at 87. Roland Rudd founded Finsbury, sold it to WPP in 2001, and chairs its distant descendant today, twenty-five years on. Earn-outs in this trade run, on most advisers’ reckoning, somewhere between one and five years. Some founders stay a great deal longer than any contract could hold them. For them the soft verb was simply true. ## The exception I could not find I looked, in all of this, for a single agency that broke ranks and announced its own sale in the plain word, on the day, in its own voice. I did not find one. I will not tell you none exists, only that I cannot remember reading it. The nearest I came was the report of a firm with no outside buyer at all, an internal succession, and even there the word sale never appeared. It was an ownership change. Even with no buyer in the room, the reflex held. None of this is a charge against the founders. The soft verb does honest work. Staff and clients are owed reassurance, and panic is expensive. Sometimes it is even the more exact word: a founder who takes part of the price in the buyer’s shares, and a partnership in the enlarged firm, really has joined something as well as sold something. The release merely declines to mention the second half. And a founder who leaves the day the covenant lapses is no traitor. That is someone whose contract has ended, behaving exactly as the contract designed them to. The euphemism is not the danger. ## The danger of believing it The danger is the founder who comes to believe his own release. Who reads the announcement, decides that yes, this is a partnership, a next chapter, business continuing as usual, and is genuinely surprised in year two to discover that it is a job, that the clients he thought were his are now assets on someone else’s balance sheet, and that the clock his announcement so soothingly left out was running for him too. ## What a founder can actually do about it Three things, and none of them costs anything. First, treat the words as a term of the deal, not a courtesy to be sorted the night before. Who drafts the announcement, who approves it, which verb it uses, and above all when your staff and your key clients hear it, and from whom: settle all of that before you sign, in the room where the money is being discussed, while you still have leverage. It sits beside the point I made [about the name of the firm](https://timosutton.substack.com/p/who-was-shandwick): if you care how the thing is said, you make it a term, not a hope. Second, tell the clients who matter yourself, and first. What a good client resents is almost never the sale. It is finding out about it from a group email, or worse, from a competitor who read the trade press before they did. Ten phone calls, made the moment you are allowed to make them, are worth more than any sentence in the release. (Where the buyer is listed, the lawyers will have views on when that moment is. One more reason to settle it before you sign.) Third, and this is the cheapest diligence there is, read your buyer’s old releases before you become one. Take the last five agencies it bought. Find the joining announcement for each. Then find where each of those founders is now, and go back and read the leaving release, if there was one, and notice which words it used and how long after the first. An afternoon of reading will tell you more about what your next few years actually hold than any number the buyer puts in front of you. ## A line for the buyer I have sat on that side too, so let me send the buyer a note rather than an accusation. If the deal you are announcing only works so long as the word sale is kept out of the room, you have quietly told yourself what you bought. Better to write a first release you will still be able to stand behind on the day you have to write the second one. ## Back to the beginning Which brings me back to Margi Booth, because her arc is the whole argument in one firm. In 2009 the deal was an important step in our evolution. Four years later a new chief executive arrived and she moved, in the [trade reporter’s phrase](https://web.archive.org/web/20200813122546/https:/www.provokemedia.com/latest/article/m-booth-hires-new-ceo-as-founder-moves-to-chairman-role), into a chairman role; we also want, she said then, to become more global, and to grow and flourish there needs to be an infusion of new ideas. And then, more than a decade on, in her own voice and unprompted, the plain word at last: when I sold. The word was always going to arrive. It just waited, as it always does, until the announcement had finished its work. --- The Sutton Memo is a newsletter on reputation and PR-agency deals, from someone who has spent three decades on both sides: crisis adviser to boards, and M&A adviser to agency founders and buyers. Some weeks it is the crisis. Some weeks it is the deal. Tim Sutton advises the owners of communications agencies on their sale, and boards in a crisis. [timsuttonpr.com](https://timsuttonpr.com). --- # Who diligences the buyer? _You will answer several hundred questions before you sell your agency. Then you will lend your buyer half the price, often more, and ask almost nothing._ Date: 2026-09-21 URL: https://timosutton.substack.com/p/who-diligences-the-buyer Anyone who has been through the sale of a business remembers the request list. The leases. The client contracts. The holiday accrual. The staff attrition table going back three years, the insurance schedule nobody has read since the day it was signed. On one agency deal I was close to, the buyer’s diligence team asked whether five of the firm’s tier-one journalists might be put on the phone to them directly, to confirm the quality of the relationship. Somebody had to explain, kindly, that journalists do not take that call. Many of them do not take ours. It is thorough. What is worth noticing is the direction. Every one of those questions runs the same way. So here is the claim, and it is the whole essay. An earn-out is a loan made without a credit check. From the morning of completion you are a lender to your buyer, usually its most junior one, and the only lender it has who never looked at its accounts. The price of skipping the check is not usually that the buyer goes bust. Sound buyers pay, and most buyers are sound. It is that when something goes wrong, and across three to five years something usually does, you renegotiate from a position of no security, against a borrower who keeps the score. I should say where I am standing. I have sat on the buyer’s side of nine agency acquisitions: seven between about 2000 and 2013, in my holding-company years, and two more since, for other buyers. Not one of those nine sellers asked a hard question about the buyer’s finances, its debt, or its record of paying the founders it had bought from before. Not one. It did not strike me as odd at the time. It strikes me as remarkable now. ## The arithmetic of the promise Start with how much of the price is a promise. Take a deal I have written about before. When Next Fifteen bought the New York agency M Booth in August 2009, the [announcement to its shareholders](https://www.investegate.co.uk/announcement/bzw/next-15-group--nfg/acquisition/1761618) was plain: initial consideration of $4 million in cash at completion, and deferred consideration of up to a maximum of $13.25 million over the following four years. More than three quarters of the top price was a promise. Go back further and the promise is still about half. When Incepta bought The Red Consultancy in 2001, the initial payment was [reported](https://www.provokemedia.com/latest/article/incepta-buys-u-k-consumer-consultancy-red) at £12 million against a maximum of £25 million. That is the ordinary shape of a deal in this trade: half the top price or more is a promise, and that has not moved much in twenty-five years. Now put the two lenders side by side. The bank lending to your buyer has, in every deal I have been near, taken security over something, written covenants it can test, and required management accounts every month. Its money is priced, dated and watched. Yours is none of those things. It carries no interest, it has no fixed date, it is payable only if a number is hit, and the number is calculated by the person who owes it to you. On completion day you are the cheapest and most patient source of finance your buyer has ever found, and it found you on the morning you were most pleased with yourself. ## When the promise met the balance sheet The fullest case I know is Australian, and it is now old enough to discuss calmly. I choose it because Photon was a roll-up, a company built by buying one business after another, which is how a good many of today’s buyers are built. Photon Group spent the 2000s buying marketing and communications businesses, largely on deferred terms. On 28 June 2010 the Sydney trade press [set out the position](https://mumbrella.com.au/photon-100m-plus-write-off-owes-banks-271m-chairman-tim-hughes-quits-earn-outs-to-total-176m-29047): the group owed its banks A$271 million, and the earn-outs still owed to the founders it had bought from were put at A$176 million. Its executive chairman stepped down the same day. Over the following months the deferred consideration arrangements were restructured. On the reporting available, the profit targets that decided the remaining tranches were cut, and part of what was owed was converted into Photon shares held in escrow. By early October 2010 Photon’s shares had lost something close to nine tenths of their value. By November 2011 the group had sold its field marketing and retail division for A$146.5 million, was [effectively clear of its bank debt](https://mumbrella.com.au/photon-group-wipes-out-debts-by-selling-field-marketing-and-retail-division-for-146-5m-63372), had shrunk from forty-five businesses to seventeen, and had A$15 million set aside to meet what the report called its “renegotiated earnout commitments”. In July 2012 it [changed its name](https://mumbrella.com.au/final-farewell-to-photon-as-the-company-becomes-enero-100592) to Enero. The easy version of this story is the wrong one. Photon’s sellers were not wiped out. They were renegotiated. The group survived and cleared its debts. The A$176 million and the A$15 million are not two measurements of the same thing on the same day, and anyone who sets them beside each other as a loss is inventing a number. The London end of it is Frank. Graham Goodkind and Andrew Bloch sold Frank PR to Photon in October 2007. In 2012, once the earn-out had been renegotiated, they and their managing director bought back a quarter of their own agency. It worked out for them, and I have written [elsewhere about what Frank did next](https://timosutton.substack.com/p/the-best-trade-in-pr). I do not know the terms and will not guess at them. The general point needs no inside knowledge: a founder in that position has stopped being somebody who is paid and become somebody who is negotiated with, and in my experience very few sale agreements say in advance which of the two you will be. ## When the promise was paper The second case involves no earn-out and no default, which is why I like it. S4 Capital built itself by merging agencies in rather than buying them on earn-outs. Announcing a new facility on 3 August 2021, it told the market the money [would provide](https://www.investegate.co.uk/announcement/rns/s4-capital--sfor/s4-capital-launches-media-monks/6819643) “approximately £200 million for general corporate purposes, including funding the cash element of future mergers, which is typically one-half of overall consideration”. The other half was S4 shares. Anybody can look up what those have done since, and I would rather you did than took my word for it. Every investor who bought in the market took the same ride, having chosen the share as an investment. The founder who merged in took it as payment. It was counted as money on the day, and it turned out to be an opinion. Cash at completion is the only part of the price that is not an opinion about the future. ## The case against all of that Now let me argue against myself, because the honest version of this essay has to. Most earn-outs owed by sound buyers are paid, and the counter-case sits inside the arithmetic I used above. M Booth’s deferred consideration, three quarters of the headline price, was by every sign honoured, and its founder stayed to run the agency through all four years of it and a year beyond. Most deals in this industry end that way, quietly, with nobody writing anything about it. Nobody has ever been delighted to announce that an earn-out was paid on time. Which brings me to my own record, and I would rather put it in myself than have somebody else find it. Interpublic was the parent of the firm I worked for. Its own filings with the SEC record that on [7 March 2003 Standard and Poor’s cut its credit rating to BB+](https://www.sec.gov/Archives/edgar/data/0000051644/000104746903032297/a2119609zs-3.htm), below investment grade, and that Fitch did the same on 14 May. Eight years later, in [a presentation to investors on 28 July 2011](https://www.sec.gov/Archives/edgar/data/0000051644/000090342311000374/ipg8k-ex992_0802.htm), the company told the market that “Moody’s took us up two notches, to investment grade, where Fitch already had us”, with Standard and Poor’s still one notch below. So from March 2003 until at least the middle of 2011, at least one of the major rating agencies had the parent company below investment grade. That covers most of the stretch in which those seven acquisitions fell. Not one seller raised it with me. Every one of them was paid what the agreement said they were owed, when it said they would be. That is the real reason nobody asks. Nobody was being lazy. For twenty-five years the question never needed asking, because the buyers were enormous and they paid. The habit was formed in conditions that no longer describe the buyer you are likely to meet. I have [already described](https://timosutton.substack.com/p/who-is-buying-agencies-now-and-why) who that buyer now is: the platform, usually private-equity-backed, occasionally listed, built to buy, integrate and be sold on, quite possibly before your earn-out ends. ## The number you no longer control In any case, default was never the commonest way to lose an earn-out. The commonest way is that the number is not hit. SRS Acquiom, which handles the escrow and payment machinery on a large volume of American deals, [publishes what it sees](https://www.srsacquiom.com/our-insights/ma-earnout-milestone-trends/), and what it sees is that “closer to one out of five dollars gets paid across all deals with an earnout”. Read that with its caveats: private-target deals generally rather than agencies, American, and from a firm whose business is the plumbing of deferred payments. But the direction is not in doubt, and it points away from the balance sheet and towards the scoreboard, because after completion you no longer control most of what decides the number. Budgets you no longer set. Hires you wanted and did not get. Central charges that land in your profit and loss for the first time. An accounting policy that changes in year two. None of that requires anybody to behave badly. It only requires the buyer to run its business as it sees fit, which is what it paid for the right to do. That is why only three of the ten questions below are about whether your buyer can pay. The rest are about whether it does, who keeps the score, and what protects you if it does not. ## Manners, and the ones who stayed silent Two more objections, and I take both seriously. Asking these questions during a courtship feels like bad manners, and a seller with one suitor has little leverage to be rude with. This is a trade in which people would sooner lose a seven-figure sum than ask an awkward question over lunch. And there is a survivorship problem in all of the above: the founders who sold to Photon and came out whole are not writing essays either. On manners: in a process with more than one bidder the questions cost you nothing, which is one more reason to have more than one bidder. With a single suitor they cost a little nerve and matter rather more, because there is no rival offer to tell you what the promise is worth. Ask them courteously and in writing. A buyer who means to pay has no reason to mind, and a buyer who will not answer has answered. On survivorship I concede the point. The founders who were paid without fuss are the majority, and they are silent. A credit check is insurance against the minority, and it costs an afternoon. ## The number hidden in plain sight Here is the part I find strange. If your buyer is listed, it already publishes what it owes to people like you, and almost none of the people it owes have ever looked. In its [accounts for the year to the end of December 2025](https://www.sec.gov/Archives/edgar/data/806968/000162828026019643/wpp-20251231.htm), WPP records contingent consideration of £39 million at fair value, with maximum potential future payments under all such agreements of up to £414 million, against closing adjusted net debt of £2,167 million. Publicis reports €305 million going out of the door on earnouts and buyouts during 2025, and a net cash position of €548 million at the year end. None of those figures should worry anybody, and I chose them because they do not. I offer no opinion on anybody’s balance sheet, which would rather defeat the purpose. They are the ordinary disclosures of large listed companies, and that is the point. The information is printed, dated and free. It tells you what your buyer owes to founders, when, and what it has to pay with. And the founders themselves have mostly never read it. The buyer you are likelier to meet now is private, and the exercise is no harder. A UK company files accounts at Companies House, and anyone can read them for nothing. The register also carries a charges section, which lists the security lenders have registered against the company and whether each charge has been satisfied, and a record of the people with significant control, which tells you who is behind the entity signing your agreement. An afternoon gets you the cash, the creditors, the shape of the borrowing and the names. I said in [an earlier piece](https://timosutton.substack.com/p/delighted-to-announce) that reading your buyer’s old announcements is the cheapest diligence there is. This is the next cheapest, and unlike the announcements, nobody wrote it to be admired. If the company signing your agreement turns out to have been incorporated last month, with a parent registered somewhere sunnier, the afternoon has given you question one. ## Ten questions to serve on your buyer So: a request list of your own. It is shorter than the buyer’s by a factor of about thirty, and it takes the form every bank has used for a very long time, because a bank deciding whether to lend asks three things. Can the borrower pay. Does the borrower pay. And what protects me if it turns out that it does not. Serve it before you grant exclusivity, while there is still somebody else in the room. Can it pay 1. Which company is actually paying me, and what stands behind it if that company cannot? 2. May I see your last three years of accounts and a note of what you currently borrow? 3. How much do you already owe to founders you have bought from, when does it fall due, and what will pay it? Does it pay 4. Of your last ten acquisitions, how many earn-outs paid in full and on time, how many were renegotiated, and were any disputed? 5. May I speak to three founders you have bought from, including one who has since left, and may I choose which three? 6. After completion, who sets my budget, signs off my hires, decides which central costs land in my profit and loss, and calculates the number my earn-out is measured on? What protects me if it does not 7. Where does what you owe me sit against what you owe your lenders, and can your lenders stop you paying me? 8. What happens to my earn-out if you are sold, merged or refinanced before it ends? 9. If part of the price is your shares: how long is the lock-up, who may I sell to when it ends, and what did the shares issued to your last three sellers turn out to be worth? 10. Will you secure, escrow or guarantee any part of the deferred price? If the answer is no, what is that part honestly worth to me today? ## What to do with the answers Three things follow from the answers, and they matter more than the questions. Price the deferred part as the loan it is. Start from what you honestly expect to earn rather than the maximum, then ask what the borrower’s record does to that figure. From a buyer who answers question four well, not much. From one who cannot answer it, a good deal. Say so out loud in the negotiation, because that discount is the only leverage the arithmetic gives you. Where the answers are thin, trade headline price for cash at completion. A lower number you have received beats a higher number you are owed, and founders who have been through it never need this explained twice. And ask for information rights running the whole life of the earn-out: the management accounts, the allocations, the definitions, in writing, every quarter. A lender who cannot see the books is a donor with paperwork. ## A line for the buyer, and a last one for you I have sat on that side of the table, so let me put this to the buyers reading rather than at them. If you have a clean record of paying earn-outs in full and on time, you are holding the cheapest competitive advantage available in a contested process, and almost nobody uses it. Build the pack: three years of accounts, the last ten earn-outs and what happened to each, and the names of three founders willing to take the call. Hand it over before anybody asks. It costs you an afternoon, it prices your promise nearer to par than your rivals can manage, and the firms that will not do it are the ones you are bidding against. [I wrote recently](https://timosutton.substack.com/p/the-best-trade-in-pr) that a holding company is really renting the founder, and that the earn-out is the term of the lease. The thought I did not follow far enough is what landlords do before they hand over a set of keys. They take references. They take a deposit. They ask, without embarrassment, to see three months of bank statements. Then they let a one-bedroom flat for a year, and keep the right to inspect it. You are proposing to lend half the value of the firm you built, possibly more, for up to five years, unsecured, to somebody whose accounts you have not opened. Your buyer will ask several hundred questions to establish whether you are good for the money. Ten of your own are the same question pointed the other way, and of the two of you, only one will spend the next five years waiting for the answer. --- The Sutton Memo is a newsletter on reputation and PR-agency deals, from someone who has spent three decades on both sides: crisis adviser to boards, and M&A adviser to agency founders and buyers. Some weeks it is the crisis. Some weeks it is the deal. Tim Sutton advises the owners of communications agencies on their sale, and boards in a crisis. [timsuttonpr.com](https://timsuttonpr.com). --- # Your regulator is having a worse year than you _The referee is playing its own match this year. Your case is one of the moves._ Date: 2026-09-23 URL: https://timosutton.substack.com/p/your-regulator-is-having-a-worse The letter arrives. An information request, a provisional decision, a notice that an investigation has been opened. The board does what boards do. Counsel is called. Exposure is estimated. Someone asks whether it is disclosable, someone else asks about the statement, and by the end of the afternoon there is a working group with a name. In thirty years of sitting in those rooms I have almost never heard the next question asked. How is the regulator’s year going? It sounds flippant. But it is the most practical question on the table, and it is the one the regulator’s own staff are asking each other in the lift. The received view of a regulator is the referee: appointed by somebody else, indifferent to the crowd, applying a rulebook that was there before the match and will be there after it. Well, yes. To a point. But the referee is playing a second match, against its owners, and lately it has been going badly. It has a budget to defend every year to people who would prefer it smaller, a chief executive whose contract has a date on it, a select committee that summons it to explain itself and, since the spring of 2025, a government asking in public whether it needs to exist at all. No institution is neutral about its own survival. A regulator that is fighting for its budget, its powers or its life does not read your file the way it read it three years ago. It reads it as a move in the other match. ## A bad year for referees Consider the weather. In December 2024 the Prime Minister [wrote to seventeen regulators](https://www.gov.uk/government/publications/a-new-approach-to-ensure-regulators-and-regulation-support-growth/new-approach-to-ensure-regulators-and-regulation-support-growth-html) asking what each would do to support growth. In March 2025 the Treasury published an [action plan](https://www.hoganlovells.com/en/publications/the-uk-governments-action-plan-for-regulators-a-new-approach) committing to cut the administrative cost of regulation by a quarter over the Parliament and to reduce the number of regulators, of which, by the government’s own count, there were well over a hundred. That same month it announced the [abolition of the Payment Systems Regulator](https://www.gov.uk/government/news/regulator-axed-as-red-tape-is-slashed-to-boost-growth), and of [NHS England](https://www.gov.uk/government/speeches/pm-remarks-on-the-fundamental-reform-of-the-british-state-13-march-2025), which is not a regulator but made the point that size was no defence. In July, at Mansion House, the Chancellor told the City that [“in too many areas, regulation still acts as a boot on the neck of businesses”](https://www.cityam.com/regulation-is-a-boot-on-the-neck-of-businesses-rachel-reeves-to-warn-at-mansion-house/). Departments have since been [told to justify](https://www.gov.uk/government/news/hundreds-of-quangos-to-be-examined-for-potential-closure-as-government-takes-back-control) every arm’s-length body they sponsor or see it closed, merged or brought back in-house. If you sit on the board of a regulated business, that is not background. That is the state of mind of the person marking your homework. None of it is new, either. In the 1980s, working for an airline trying to get onto European routes, I watched regulators who had resisted the argument for competition for years move the moment their owners changed what they wanted. Then the instruction was competition. Today it is growth. The rulebook moves when the referee’s position moves. There are, as far as I can see, four ways for a regulator to be in trouble, and they produce four different regulators. Fighting for its life. Flinching from its owners. On probation. And fighting somebody else’s war. Some are in two of these at once, which is a comfort to nobody. Each has a name this year, so let me take them in turn. ## Fighting for its life: Ofwat In July 2025 Sir Jon Cunliffe’s [Independent Water Commission](https://www.gov.uk/government/publications/independent-water-commission-review-of-the-water-sector) delivered 88 recommendations and called for a “fundamental reset” of the water industry, among them that Ofwat should be abolished and its work folded, with the Drinking Water Inspectorate and the water functions of the Environment Agency, into a single regulator. The Environment Secretary accepted the central recommendation in the Commons before the day was out. A [white paper](https://commonslibrary.parliament.uk/water-reform-a-new-vision-for-water/) followed in January, promising a transition plan and a Water Reform Bill. As I write, neither has appeared, no timetable has been confirmed, and the only bill on Parliament’s books to abolish Ofwat is a [Liberal Democrat presentation bill](https://bills.parliament.uk/bills/4000) that has had its first reading and nothing else. Ofwat’s chief executive [stood down](https://www.ofwat.gov.uk/chief-executive-ofwat-david-black-to-stand-down/) a fortnight after the decision and the post has been held on an interim basis since. Ofwat’s own response to the white paper was that a new regulator would “renew focus, improve the sector for customers, investors and the environment, and rebuild trust”. It is not every day you see the condemned body welcome the sentence. So Ofwat exists, has its powers, is taking decisions, and has a death warrant with no date on it. Its biggest case is still open. Thames Water, which Ofwat [proposed to fine £104 million](https://www.ofwat.gov.uk/thames-yorkshire-and-northumbrian-water-face-168-million-penalty-following-sewage-investigation/) in August 2024 as part of a record £168 million tranche across three companies, is now the subject of a [Commons committee report](https://committees.parliament.uk/committee/52/environment-food-and-rural-affairs-committee/news/217944/) of 18 September urging the government either to put it into special administration or to take emergency control of its finances before the money runs out at the end of the year. ## Tough to the end, or waiting it out What does an enforcement decision look like when the enforcer is being wound up? I have seen two versions, and a board needs to know which one it is facing. The first is the body that wants the record to show it was always tough: its final year produces the largest fines, the firmest language and the least room. The second is the body that has quietly stopped taking any decision its successor might have to own, so that everything difficult drifts into the transition. Both are rational. Neither is about you. From outside you cannot yet tell which Ofwat is. From inside, with a live file, you can, by the simplest of tests: has it taken a contested decision since July 2025 that its successor will have to defend, and how long did it take? And the second version has a consequence boards routinely miss: the counterparty you are actually negotiating with is not the regulator across the table but the one that does not yet exist, and behind it the department writing its terms of reference. If you are Thames Water, or any of the other companies waiting on penalties and price reviews, your real regulator this year is a white paper. ## Flinching from its owners: the FCA The Financial Conduct Authority’s last two years are a lesson in what happens when a regulator’s own overseer turns on it. In February 2024 it proposed to name the firms it was investigating, replacing a presumption of privacy with a public interest test. The City objected, which was to be expected. Then the Treasury objected, and the House of Lords committee that oversees financial regulation objected, and by November the proposal had shrunk. On 11 March 2025 the FCA [dropped it](https://www.cliffordchance.com/insights/resources/blogs/regulatory-investigations-financial-crime-insights/2025/03/fca-drops-name-and-shame-proposals-for-regulated-firms.html). A month later its chief executive was [reappointed](https://www.fca.org.uk/news/press-releases/nikhil-rathi-reappointed-fca-chief-executive) to 2030, the first second term in the role’s history, and said plainly that the Treasury had let him know what it thought of the plan. Then came the motor finance redress scheme, some £7.5 billion of compensation for commissions on car loans. This was not a fight the FCA picked. The courts had all but obliged it to build the scheme once the [Supreme Court had ruled](https://www.supremecourt.uk/cases/press-summary/uksc-2024-0159) on the commissions in August 2025. The FCA [confirmed the scheme](https://www.fca.org.uk/news/statements/fca-confirms-motor-finance-redress-scheme) in March. By July the Upper Tribunal had [suspended parts of it](https://www.fca.org.uk/news/statements/motor-finance-scheme-partially-suspended), on terms the FCA agreed with the four parties challenging it, three of them lenders, until the case is heard in December at the earliest. A regulator that has been told to promote growth, slapped down for wanting to be more transparent, and then obliged to pause the one big case it did not choose, all inside two years, is not the regulator of 2023. ## Caution is not your victory Here is the mistake boards make with a flinching regulator. They read its caution as their victory. It is nothing of the kind. A regulator that has been told to pick fewer fights picks the ones it is certain to win, and it needs to win them, because a lost case is now an argument for its abolition. The file that survives that filter is pursued harder, not softer. And a regulator talked out of naming the firms it investigates still investigates them; it has been denied the cheap deterrent it wanted and will look for another. Do not mistake a regulator’s political weakness for weakness on your file. The two are more often inversely related. ## On probation: the CQC, Ofsted and the SRA The third condition is the regulator that has been declared unfit in public and told to rebuild itself while carrying on regulating. In July 2024 Dr Penny Dash’s [review of the Care Quality Commission](https://www.gov.uk/government/publications/review-into-the-operational-effectiveness-of-the-care-quality-commission/review-into-the-operational-effectiveness-of-the-care-quality-commission-interim-report) found inspection levels below where they had stood before the pandemic, judgments that varied from inspector to inspector, and an IT system that did not work; the Health Secretary called the body [“not fit for purpose”](https://www.communitycare.co.uk/2024/07/26/cqc-not-fit-for-purpose-says-streeting-in-wake-of-damning-report/). A new chief executive arrived that December. The backlog of unpublished reports has since been cut from about five hundred to a handful, on the [CQC’s own account](https://www.cqc.org.uk/about-us/improving-how-we-work/1225-update), and the assessment framework it introduced in 2023 is being rewritten. Ofsted, after the coroner’s findings on the death of the headteacher Ruth Perry, [abolished single-word judgments](https://educationhub.blog.gov.uk/2024/09/02/removal-ofsted-single-word-judgements-schools/) in September 2024 and introduced report cards a year later. When they were first proposed, Mrs Perry’s sister, Professor Julia Waters, [called the plan](https://www.itv.com/news/2025-02-02/fears-that-ofsteds-proposed-grading-scale-is-rehash-of-dangerous-system) “a rehash of the discredited and dangerous system it is meant to replace”. And the Solicitors Regulation Authority, which exists to hold solicitors to account, is being held to account by the Legal Services Board, which exists to hold it to account, a sentence only the legal profession could have arranged. The LSB issued formal directions over the SRA’s handling of the collapse of Axiom Ince. A second firm, PM Law, fell in February, taking the client money lost across the two to [about £100 million](https://legalservicesboard.org.uk/news/legal-services-board-statement-on-sra-regulatory-performance-6-may-2026), and the LSB has now asked for an independent audit of whether the SRA did what it was told. ## Your inspection is the inspector’s exam A regulator on probation behaves in a knowable way. It is process-heavy, because process is what its overseer can count. It is slow, because it is checking its own work. And it is hungry for a visible win, because it needs to show the rebuild has taken. If you are the regulated party, your inspection is also the inspector’s exam. Which suggests something that sounds cynical and is not: the provider that helps a rebuilding regulator look competent, with clean submissions, no ambushes and no surprises in the press, is trading in the only currency the regulator wants that year. That is not gaming the system. It is reading the room. Note what is on that list and what is not. None of it is a favour, and none of it is flattery, which inspectors can smell. All of it is what a well-run provider does anyway. The difference this year is that somebody is counting. ## Somebody else’s war: Ofcom The fourth condition is the strangest, and Ofcom is living in it. For most of its life Ofcom was argued with by British broadcasters and British telecoms companies. Then the Online Safety Act handed it the American internet, with the [first duties in force](https://www.ofcom.org.uk/online-safety/illegal-and-harmful-content/enforcing-the-online-safety-act-platforms-must-start-tackling-illegal-material-from-today) from March 2025. This spring it fined the message board 4chan £520,000, and 4chan [declined to pay and sued Ofcom](https://www.biometricupdate.com/202602/ofcom-fines-kick-threatens-4chan-as-osa-enforcement-steadily-dials-up) in a federal court in the United States, where the Act has few friends. It is a novel position for a British regulator: sued by the party it fined, in a court three thousand miles from anyone who has to obey it. Closer to home, in February 2025, it lost its first judicial review to a broadcaster, when the High Court found its impartiality rulings against GB News were [“vitiated by error of law”](https://www.judiciary.uk/judgments/gb-news-v-ofcom/); within a fortnight Ofcom had [withdrawn every remaining investigation](https://www.mishcon.com/news/high-court-quashes-ofcoms-decisions-against-gb-news) into politicians presenting programmes. Its chair [changed in June](https://www.gov.uk/government/news/sir-ian-cheshire-confirmed-as-new-ofcom-chair). What this means for a British broadcaster or telecoms company with a case in front of Ofcom is uncomfortable. Your case is being decided by a body that is a defendant abroad and has recently been told by a judge that it misread its own code. A regulator in that position does two things at once. It over-corrects toward caution wherever the last loss was, hence the withdrawn investigations. And its most visible enforcement so far has landed on the least sympathetic target on the internet, which, whatever the reason for the choice, is the easiest kind of win to explain at home. If your case sits anywhere near the fault line, it is read against the bigger fight, not on its merits alone. When the regulator is a proxy, so is everyone in its queue. ## What to do about it Three things, none of them glamorous. First, read the regulator’s papers with the attention you give your own. Its annual report and its risk register. Its board minutes, which most of them now publish. The transcript of its last appearance before a select committee. The letters between it and its sponsoring department, which are usually on the department’s website and almost never read by anybody outside it. Half of it is dull. The other half tells you what the regulator is afraid of this year, and what it is afraid of is what it will act on. Second, put its calendar beside yours. A decision due in the month a regulator is defending its budget, awaiting a bill or waiting for a new chair to be confirmed is a different decision from the same one taken in an ordinary month. I am not suggesting you time your submissions to catch the referee looking the other way. Regulators are not stupid, and the ones under pressure are the least stupid of all. I am suggesting you understand the answer you get, and do not mistake a cautious one for a favourable one, or a fierce one for a personal one. Third, and this is the one boards most need to hear, never argue the regulator’s weakness in public. The temptation is real. It has just lost in court, or been called unfit by its own minister, or been told it is to be abolished, and your case is strong. Saying so converts a dispute about your facts into a dispute about its existence, and it will fight that one with resources you do not have and with a public and a press that will take its side against a company nine times in ten. Make the point in private, where it may even land. In public, be the party in the room that behaved like a grown-up while the regulator was having its bad year. That is an asset, and it is banked with the successor body as well as the current one. ## In fairness to the referee I should concede two weaknesses in all of this. The first is that none of it is a complaint about the people. Most regulators I have dealt with are staffed by people who could earn twice as much on my side of the table and have chosen not to, and they know their institution’s position far better than any board does. The second is that not every regulator is in trouble, and it is lazy to assume so. Ofgem’s year, as far as I can tell, has been a year of price-cap methodology and heat networks, which is what an ordinary regulatory year looks like, and I mean that as a compliment. Plenty of regulators are simply doing their jobs, and a board that treats every one of them as a wounded animal will misread the healthy ones as badly as the cynics misread the sick. The discipline is not to assume. It is to check. But do check. When the letter comes, the first question is still what did we do. The second, asked far less often, is what is happening to them. The board that asks both is not being cynical. It is being the only party in the room that has read all the papers. Your regulator may well be having a worse year than you. Its own staff already know. It would be a pity to be the last person in the building to find out. The Sutton Memo is a newsletter on reputation and PR-agency deals, from someone who has spent three decades on both sides: crisis adviser to boards, and M&A adviser to agency founders and buyers. Some weeks it is the crisis. Some weeks it is the deal. Tim Sutton advises the owners of communications agencies on their sale, and boards in a crisis. [timsuttonpr.com](https://timsuttonpr.com) --- # The year after the good year _The best year in your agency's history is the worst year to sell it in._ Date: 2026-09-25 URL: https://timosutton.substack.com/p/the-year-after-the-good-year Every founder of a growing agency carries two numbers around. The first is in the last set of signed-off accounts: last year’s profit, some months out of date and faintly embarrassing, because the business has moved on since. The second is the number the founder actually believes in: this year’s profit, watched monthly, the run-rate. It is bigger. It is always bigger. That is how you tell them apart. A buyer’s model starts with the first number. When I was [buying agencies](https://timosutton.substack.com/p/what-i-learned-buying-agencies) for Weber Shandwick, we would often take the last two audited years, add the current year as projected, and average the three. There is nothing sinister in that. A finished year is the only kind anybody can check, and a founder sitting in the buyer’s chair would do exactly the same. But look at what the arithmetic does to a business that is growing fast. The year the founder is proudest of gets one vote in three, and the two years before it outvote it. So the faster the growth, the further the buyer’s number sits below the founder’s. Negotiation does not close the gap so much as move it, out of the money paid on completion and into the money the founder has to earn afterwards. ## The number the buyer pays on Put some numbers on it. I will use index numbers rather than pounds, so that nobody reads their own business into them. Take an agency whose profit went from 100 to 150 and is heading for 225 this year. Its founder, quite reasonably, thinks of it as a 225 business. The average of the three years is 158. Now take a steadier agency that has made 200 in each of the same three years. Its average is 200. The steady business is valued on the number its owner believes. The fast one, which is making more money this year than the steady one, is valued on less: on a number it passed some time last year. A good buyer will pay something for the growth, usually through a higher multiple. The premium is real and I do not want to pretend otherwise. But the multiple is applied to the average, and a generous multiple of a diluted number is still a diluted price. The rest of the founder’s growth is deferred and made conditional, which is not the same thing as being paid. ## Paid for the past, renting the future You can see the shape in the buyers’ own announcements. When Next 15 bought Shopper Media Group in 2021, [its announcement](https://www.investegate.co.uk/announcement/bzw/next-15-group--nfg/acquisition-of-shopper-media-group-ltd/6449345) set out an initial consideration; then a further payment due around February 2022, based on the business’s EBITDA for the year to September 2021, the year then under way; then further payments around June 2023 and June 2025, based on performance in the two-year periods ending in January of those years. Read that as a founder would. The years already in the accounts were paid for on the day. The year in progress was paid for once it had closed and been counted. The years after that were paid for years later, if they happened. That is what paying for growth means in practice. The future a buyer pays for most readily is one that has already happened. None of this is peculiar to one buyer. It is simply how the thing is done, and I have [written before](https://timosutton.substack.com/p/who-diligences-the-buyer) about what the deferred money really is, which is a loan from the seller to the buyer. The point here is narrower. The size of that loan is set by the gap between the two numbers, and the fast grower has the widest gap there is. ## Growth eats cash Growth costs money before it makes any. The new team is paid from its first month. The new client pays when its procurement department gets round to it. Work in progress builds, debtors swell, and for a while the bank balance of an agency that is winning everything can look very like the bank balance of one in trouble. A buyer notices. Profit still sitting in the debtors’ ledger looks, to the [analyst who takes your claims apart](https://timosutton.substack.com/p/the-analyst-who-takes-your-claims), less like profit and more like a promise. How much of last year’s profit turned into cash is one of the first questions asked, and a business that is growing fast gives the weakest answer it will ever give. Then there is completion, and an argument about cash. The seller wants to take the spare cash out before the sale, which is reasonable, since it is theirs. The buyer wants enough working capital left in the business that it will never have to put in money of its own. Most deals value the business as if it came with a normal level of working capital, and settle the difference afterwards, once the completion accounts are agreed. If the business falls short, the deficit “will be deducted from the purchase price”, in the words of [one accountancy firm’s guide](https://www.hwfisher.co.uk/normalised-working-capital-and-completion-mechanisms/), which also calls agreeing the normal level, rather gently, “a subjective area”. ## The compromise The compromise I usually saw was sixty days of net working capital, counting both the money due in and the bills due out, measured on the business as it operates now. Leave more than that in and you are paid for the excess, pound for pound; leave less and it comes off the price. Fair enough, on its face. But notice which business the sixty days is measured on. The price was set on an average of three years. The working capital that has to be left in for that price to stand is set on the business at its current size, because that is the size the buyer has to fund from the Monday after completion. For a steady agency those are the same business. For a fast grower, the buyer values the smaller agency and asks for the working capital to run the larger one, and the seller will be arguing for the lower figure with someone who has had this argument many times before. ## New revenue has not been tested A client won nine months ago has not renewed once. It may stay for a decade. It may put the account out to pitch next spring. Nobody knows, least of all the buyer, and a buyer values what it cannot yet know at a discount. The same goes for people. A hire made six months ago has no record in the business, however good the CV. And growth hides concentration rather well. The two big wins that made the year are also the two clients whose loss would unmake it. Then margin. An agency that is growing hard is usually hiring ahead of the revenue it is chasing, which is the right thing to do, and which depresses margin in precisely the year the founder most wants to show off. The buyer sees revenue that has jumped and a margin that has dipped, and asks which of the two is the real trend. ## What the offer looks like As a rule of thumb, when I was buying, half the price was paid in cash on completion and half was earn-out, and the split moved according to how proven the business was. The stronger the evidence, the more of the price came on the day. Now assemble the offer the fast grower receives. A multiple, possibly a generous one, applied to an average the business has outgrown. An argument about what normal working capital means. And a cash share at the cautious end, because proof is precisely what a business in the middle of a surge cannot supply. The rest is deferred against targets set on the growth the founder already believes in. The headline may look perfectly respectable. The structure is where the fast grower pays. In effect the founder is asked to earn the same growth twice: once to build it, and again to be paid for it. The risk in the new business is the founder’s either way. Sell now and you carry it as the buyer’s employee, for the deferred part of the price. Wait, and you carry it as the owner, for all of it. ## In fairness to the buyer Some buyers will pay for the future, and I should say so. Private equity-backed platforms, in my experience, are the readiest to price on the most recent twelve months rather than a three-year average, and sometimes to count the year’s contracted wins as if they had already run for a full year. They say so, and the difference is real. Listed buyers have done it too. When M&C Saatchi bought a third of the American agency SS+K in November 2014, [the announcement](https://www.investegate.co.uk/announcement/rns/m-c-saatchi--saa/us-acquisition/3793185) put an “estimated consideration” on the stake, based on a multiple of profit “applied to 2014 and 2015 performance”, with 2015 not yet begun. But notice the word estimated. A buyer that pays for the future has not stopped discounting it. It has moved the discount from the price into the contract: an adjustment once the year is counted, a longer earn-out, a sliding scale that pays less the further short of target the business lands, tighter terms for a founder who leaves early and, with a platform, a slice of the price paid in the platform’s own shares, which is a deferral with a nicer name. The growth priced in on day one is growth the founder now has to deliver to be paid for. The risk has not gone away. It has changed address. ## Growing, or just getting bigger There is a test a founder can run tonight. A buyer will run it in the first week, so it is kinder to run it first. Headcount growth is not profit growth. An agency whose margin held while its headcount rose has grown. One whose margin fell while its headcount rose has got bigger, which is a different thing and is valued differently. The two numbers that tell you which you have are [revenue per head](https://timosutton.substack.com/p/ai-is-not-coming-for-pr-it-is-coming) and margin, watched quarter by quarter rather than admired once a year. ## Leave the buyer something to do I remember one business with a good current year where the growth was not, in the end, what excited us. What excited us was that we could see, quite clearly, how the margin could be improved once the business was inside the group. That visibility did more for our return than the growth did, and most of the benefit came to us, which at the time I thought entirely proper. The obvious lesson is to fix the margin before you sell, and it is only half right. A business in which every possible improvement has already been made is less attractive to a buyer, not more, because it has removed one of the main places the buyer’s return was going to come from. So separate the improvements into two kinds. There are the ones only the owner can make: pricing, utilisation, hiring behind revenue rather than ahead of it, the client who costs more to serve than it pays. Make those. Once they are in a finished year’s numbers they are paid for at the full multiple, and a buyer who finds them undone will simply make them and keep the proceeds. Then there are the improvements only a buyer can make: scale, a shared back office, selling across the group’s other firms, buying power the owner will never have. Leave those, visibly, and say in the sale document where they are. They are the buyer’s reason to buy. A business with nothing left to improve is a finished product, and a buyer pays its premium for what it can still add. ## The year after the good year None of this is an argument against growing, or against selling. It is an argument about timing. Go back to the index numbers and wait a year. The years in the accounts are now 150 and 225. This year’s projection is, say, 250, as the growth settles. The average is 208, almost a third higher than it was twelve months earlier, for doing nothing more heroic than letting the good year turn into a set of accounts. Meanwhile the clients it brought have renewed once, the hires have a record, the margin has recovered from the hiring, and the founder’s number and the buyer’s number have come close enough together that there is far less left to defer. Proven strength, remember, is what moves the split towards cash. I should be honest about the multiple. Settled growth is less exciting than a surge, and the multiple offered a year on may be a little lower for it. What has changed is the share of the price that arrives as money and the share that arrives as a promise. Founders who have been through an earn-out tend to regard that as a good trade. And if the growth does not settle, and the surge runs on into another good year, the same arithmetic applies to that one, and its founder has a nicer problem than most. This is not an instruction to wait forever. Growth levels off, founders tire, markets turn, and the year after the good year is a year, not a decade. Nor is it an instruction to sit by the telephone, which [has its own cost](https://timosutton.substack.com/p/the-call-comes-a-funding-cycle-late). It is a warning against selling in the middle of the surge, when the two numbers are furthest apart and the buyer can verify least. ## While you wait In the growing years, the useful work is unglamorous. Keep the figures a buyer will want, with the founder’s own drawings and the one-off costs cleanly separated. Keep a list of renewals, by client and by date, because a buyer will ask for one. Watch revenue per head every quarter. Hire behind revenue where you can bear to. And once a year, read the business as a buyer would, before anybody else does. If an offer does arrive early, and they often do, since buyers read growth charts too, two facts will tell you what the buyer really thinks of your growth: the number the multiple is applied to, and the share of the price that is cash on the day. If the first is this year’s number and the second is most of the price, the buyer is paying for the growth, and that is an offer to take seriously in any year. If the first is an average and the second is half, you are being asked to lend the buyer your own growth, and the year after should pay more of it back in cash. The good year is worth more in the accounts than in the forecast. The founder who sells in the middle of it will hear a great deal about the multiple. In all my years in the buyer’s chair I do not remember paying a multiple of a run-rate. I remember promising a few. The Sutton Memo is a newsletter on reputation and PR-agency deals, from someone who has spent three decades on both sides: crisis adviser to boards, and M&A adviser to agency founders and buyers. Some weeks it is the crisis. Some weeks it is the deal. Tim Sutton advises the owners of communications agencies on their sale, and boards in a crisis. [timsuttonpr.com](https://timsuttonpr.com) --- # If a lion could talk _We talk about AI as if it had a life. It reads us as if we were text. Wittgenstein's lion explains both mistakes._ Date: 2026-09-26 URL: https://timosutton.substack.com/p/if-a-lion-could-talk If a lion could speak, we couldn’t understand him. I had it that way for years, until I checked, and I suspect most people who quote it do too. What Wittgenstein’s translator, Elizabeth Anscombe, gave us is this: “If a lion could talk, we could not understand him.” A small correction, but worth making at the top of an essay about misunderstanding. My subject is not really the lion, or even Wittgenstein. It is the AI machine most of us now talk to every day, which has our language word-perfect and no life at all, and the two mistakes that follow from that: the one we make when we talk about it and the one it makes when it reads us. Wittgenstein’s sentence, read properly, explains both. It needs a little philosophy first, but I promise not much. The usual reading is that a lion would have strange things to say. Its concerns would be antelope and territory, ours would be mortgages and the school run, and the conversation would be heavy going. That is true but it is not what he meant. The sentence follows a passage about people, not animals. One human being, he says, can be a complete enigma to another, as we find in a strange country with entirely strange traditions, even with a mastery of the language. “We cannot find our feet with them.” The lion is the extreme case. Words take their sense from the life in which they are used: the hunting, the hunger, the fear, the pride. A lion that spoke English would hand us the words without the life that gives them their point. We would hear sentences, parse them perfectly, and understand nothing. Seventy years on, we have built the lion the other way round. ## Two Wittgensteins There are two Wittgensteins, which is one more than most philosophers manage. The young one wrote the Tractatus, published in 1921, which gave us “The limits of my language mean the limits of my world.” Language, on that view, is a picture of the world. A sentence depicts a fact, and where the sentences run out, so does the world, or at least mine. The popular explainers reach for that line when they reach for the lion, which is a pity, because the older Wittgenstein spent his later years taking it apart. In the Philosophical Investigations, published in 1953, meaning is no longer a picture. For a large class of cases, he says, a word’s meaning is its use in the language. And use happens inside what he called a form of life, the shared activities, needs and habits in which words do their work, so that to imagine a language is to imagine a form of life. When we agree in the language we use, it is “not agreement in opinions but in form of life.” On the young man’s view, or a rough version of it, a machine that had read every sentence we ever wrote would have our whole world, since the world is what the sentences picture. On the old man’s view, it would have almost none of it, because the sentences were only ever the visible part of something we do together. I am going to stand with the old man, and I shall try to say so only once. ## All of the language, none of the life The lion has a form of life and no language. A large language model has the reverse. It was trained on the written residue of nearly every human form of life, and it is a member of none. It has read more about grief than any widow, and it has never lost anyone. On Wittgenstein’s test, then, it ought to be harder to understand than the lion. It does not feel harder. It feels easier than most of the people I have sat across a table from. The lion’s silence tells the truth about the gap. The machine’s fluency covers it over. The better the prose, the more it looks like a fellow speaker, which, on the later Wittgenstein’s account, is the one thing it is not. Thomas Nagel’s bat tends to wander in here, and I should keep it out. His question of 1974, what it is like to be a bat, is about experience, and whether it has an answer for machines I do not know. The lion’s question is easier. It asks only whether we share a life with the thing and we do not. ## The wrong language or an unreadable one There is an exact description of what a language model does, and it is mathematics: billions of numerical weights, a probability distribution over the possible next pieces of text, a sample drawn from that distribution, and then the whole thing again. Tell a colleague that the model assigned high probability to a token sequence with no counterpart in its source documents, and you will lose them halfway through. Tell them it made something up, and they have it at once. So we reach for the ordinary verbs. It thinks, it knows, it wants, it understands, it reasons. Every one of them imports a life the thing does not have. That is anthropomorphism, and we stand rightly accused. But we are not careless so much as stuck, between a language that is wrong and a language nobody can read. Nor could we help it if we tried. In 1944 the psychologists Fritz Heider and Marianne Simmel showed thirty-four students a short film of two triangles and a circle moving around a box, and asked them to describe it. One described shapes. The other thirty-three described living things, mostly people and in two cases birds, and nineteen told it as a story, the one the authors chose to print being two men fighting over a girl. Twenty years later Joseph Weizenbaum wrote ELIZA, a program that mostly turned your own sentences back into questions. His secretary had watched him build it and knew precisely what it was. After a few exchanges she asked him to leave the room, so that she could carry on the conversation in private. He was dismayed. I am no better. Last month I published an essay called [AI has already read your crisis statement](https://timosutton.substack.com/p/ai-has-already-read-your-crisis-statement). Read is a lovely verb, and it is quite wrong. I have used it twice already in this essay and I shall use it again, because there is no better one to hand. ## The prediction about us Alan Turing saw the difficulty in 1950 and declined to fight it. He called the question of whether machines could think “too meaningless to deserve discussion”, swapped it for a game, and made two predictions for the end of the century. One concerned how well the machines would play his game. The other concerned us: “the use of words and general educated opinion will have altered so much that one will be able to speak of machines thinking without expecting to be contradicted.” The one about us came in roughly on time. The machines took another twenty years to catch up with the vocabulary. Ordinary language is not doomed, then. It stretches. Computer was a job title for three centuries before it was a machine: a person, latterly and usually a woman, paid to do sums by hand. Nobody now thinks a computer is a person. Thinks will go the same way, and knows, and reasons. The cost is that when thinks comes to mean what the machine does, we shall need a new word for what we do. ## The trick of the borrowed word Stretching is one thing, but some words carry a picture across with them. The people who could have given us better words are the ones who hold the mathematics, and they chose the ordinary ones too. OpenAI calls its models that work through a problem step by step before answering “reasoning models”. Anthropic offers its model “extended thinking”, which is more than most of us are offered. And the whole industry, laboratories included, has settled on hallucination. In 2020 I wrote an essay for [PRovoke](https://www.provokemedia.com/latest/article/opinion-for-data-the-end-of-the-age-of-innocence) about another word, and [came back to it last month](https://timosutton.substack.com/p/the-automation-of-judgment). I wondered whether the rot had set in when we stopped saying statistics and started saying data: everyone knew statistics were problematic, whereas data suggests objective truth. Hallucination works the same trick in the other direction. Data took the product of human choices, what to count and how to describe it, and made it sound like a reading off a machine. Hallucination takes the product of a machine and makes it sound like a human lapse. ## A confident guess A model producing a fluent falsehood is doing what it does the rest of the time, which is generating likely text. Sometimes the likely text is true and sometimes it is not, and the sampling is the same either way. To call the false ones hallucinations is to imply a mind normally in touch with reality that has briefly lost its grip. It is no compliment, but it is a courtesy: it supposes there was a grip to lose, which the research has not settled. What the training rewards is less of a mystery, and OpenAI, to its credit, says so: a confident guess beats an admission of uncertainty. That describes an examination candidate, not a visionary. The more careful word, which Geoffrey Hinton among others prefers, is confabulation. It comes from the neurology ward, where it describes patients with damaged memory who fill the gaps with fluent, sincere invention. Even the correction borrows a patient. ## The beetle in the box Here I owe the other side its best case, and it is a good one. Wittgenstein is an awkward ally for anyone who wants to say the machine lacks something inside, since much of the Investigations goes to showing that the inner something drops out of the question. His best-known illustration is a box. Suppose everyone had a box with something in it, which they called a beetle. Nobody can look into anyone else’s box, and everybody knows what a beetle is only by looking at their own. The word beetle still works well enough in conversation. So, he concludes, the thing in the box “has no place in the language-game at all; not even as a something: for the box might even be empty.” Apply that to the machine and it bites. If meaning is use, and the machine uses words as we do, fluently and in context, by what right do we say it does not mean them? Its box may be empty. For the purposes of the language game, so might ours. I find this more persuasive than I would like. The answer is Wittgenstein’s too, and it is not a beetle by another name. Use, for him, is never bare output. Words do their work in the course of doing something, buying apples, giving orders, reporting a pain, among people who share the doing. A form of life is not hidden in a box. It is the most public thing there is: bodies, meals, births, deaths. The machine has the record of all that, which is text. What it lacks is not something inside that nobody can see but something outside that anybody can. ## The octopus and the bear Emily Bender and Alexander Koller made the point in 2020 with an octopus. Two people stranded on separate islands talk through an undersea cable. An intelligent octopus taps the line, learns the patterns well enough to impersonate one of them, and does fine until the other asks for help fending off a bear with sticks. The authors put the bear question to GPT-2, a forerunner of ChatGPT, and one answer began: “Take one stick and punch the bear, and then run faster to the store.” Today’s models would probably manage the bear, having read a good deal about bears. That is rather the point. They manage by having read. The machine has learned every move in the game from the scorebooks. It is being sent onto the pitch now, with tools and errands, and it has still never been hungry. ## The reverse charge So much for our error. What is the machine’s? The word exists. In 1986 the psychologist Linnda Caporael called it mechanomorphism, seeing machine-like behaviour in other human beings. She meant a habit of ours. I am lending it to the machine, which is the trouble with read all over again, and I see no way round it. It meets us as data, text mostly, lately pictures and voices too, and none of it is happening to the machine. So it reads us as data. A letter of condolence is a passage of negative sentiment. A reputation is a cluster of words that tend to appear near a name. A joke that does not land is a classification error. None of these descriptions is exactly false. They are the view of us that a lion would have if it had read every book humans ever wrote and never met one of us. And here the two Wittgensteins come back. The machine holds no theory, and the young man’s was subtler than its slogan, but the slogan fits. For the machine, the limits of its language really are the limits of its world. Whatever it reaches for to check a claim, a search result, a document, a photograph, arrives as more of the same, and nothing in its making gives it reason to suspect there is anything else. Its misreading of us is the Tractatus, running at scale. ## We did it first It would be comfortable to leave the charge there. But it learned its picture of us from us, and the picture was there before it arrived. My own trade turned the public into sentiment scores and share of voice before the current models came along, a habit I have [argued with before](https://timosutton.substack.com/p/reputation-needs-more-than-a-sentiment). We do it to ourselves most days. We lack the bandwidth. We are hardwired to worry. We need time to process. We could do with a reboot. The vocabulary of the modern self-help shelf would serve quite well as a manual for a laptop. So the machine did not invent the reduction; reducing us to data is what it is built to do. What it took from us was the manner. The dashboards and the bandwidth were in the text before it read a word. The words it reduces us with are ours. That leaves the two charges as one. We give the machine a life it does not have, because its language is ours and we cannot hear language without hearing a life behind it. The machine takes our life for language, because language is all of us it has. Each meets the other in language, and mistakes that for meeting. The later Wittgenstein’s answer, a shared life, is the one thing neither can give the other, and no fluency on either side will supply it. Wittgenstein’s lion was only ever hypothetical. Ours exists, and answers back, at length, in perfect English. The lion, at least, had the decency not to talk. The Sutton Memo is a newsletter on reputation and PR-agency deals, from someone who has spent three decades on both sides: crisis adviser to boards, and M&A adviser to agency founders and buyers. Some weeks it is the crisis. Some weeks it is the deal. Tim Sutton advises the owners of communications agencies on their sale, and boards in a crisis. [timsuttonpr.com](https://timsuttonpr.com) ---