Business

Nobody Is Buying Hours Anymore

Services as Software

I caught up last week with a friend I have known for most of my career. He leads transformation at a large services firm, the kind of shop that lands on the shortlist for serious enterprise work. We talked for forty minutes. Five of those minutes were about our kids.

The rest was the most useful conversation I have had this year about what AI is actually doing to an industry, as opposed to what people say it is doing from a stage.

The last piece I wrote was about the inside of one firm. How we decide, what we let the machine do, and where we keep the veto. This one is about the outside. What happens to an entire industry when the thing it has always sold by the hour turns out to be the thing the machine is best at.

He said something in the first two minutes that I have not stopped turning over.

“We have spent millions in real dollars on our versions of this, and we are not making the progress we need to make.”

The sentence that ends the business model

Here is what I said back, and I did not soften it.

You cannot sell bodies by the hour anymore. Well, you can. You will just die, because nobody is buying that way.

He agreed instantly, which told me he had already been living it. His firm now runs AI right through its offerings and across its deals. The question is no longer whether AI is being used. It is whether the firm captures any of the value it creates.

And then he said the part that actually matters.

“We do not have the pricing model yet.”

Sit with that. The capability is in the building. The clients want it, and increasingly they expect it. The commercial model underneath has not moved at anything like the same speed.

That is the problem professional services firms have to solve, and it is worth stating precisely. The industry has spent decades monetising effort. AI reduces the effort required to produce the outcome. Become dramatically more efficient while pricing exactly as before, and the economic value of that efficiency flows to the client instead of accruing to the firm that created it.

His question was the right one. How are you capturing the value? How are you maintaining EBITDA? Otherwise you just gave it away.

That is the whole disruption sitting in a single line item. If you bill by the hour, every efficiency you introduce becomes a tax on your own revenue. You are investing millions to build the thing that reduces the hours you have to sell, and unless you change how you price and package the work, you never capture what you created.

Clients are already buying services the way they buy software, whether or not the firm selling to them has noticed. An outcome, a renewal, a number they can plan around. Not a rate card.

The double transformation

The spend is worth understanding, because it describes the position almost every established services company is in right now.

His firm has invested millions in AI and knows it is not moving fast enough. That does not make it unusual. It makes it representative of an industry trying to change the engine while the plane is still flying.

Because a professional services firm has two transformations running at once, and only one of them gets talked about.

The first is its own. How it sells, operates, designs, builds and delivers, all of it rebuilt around a technology that changes the unit economics of the work underneath.

The second belongs to its clients. At exactly the same moment, those same clients expect the firm to lead them through the identical transformation, using the identical technology, and to be credible while doing it.

That is the real double whammy. You take your own company through the disruption while taking your customers through it simultaneously, and you do both while the economics, the technology, client expectations and the competitive field all move underneath you.

The firms in the most trouble are not the ones that ignored AI. They are the ones that cannot move fast enough from experimentation to a genuinely different operating and commercial model.

He described the position better than I could. Too big to act small, too small to act big.

The market is changing underneath him too. Every deal is about AI now, no matter what is technically on the agenda. That is slowing some decisions down, changing deal structures, and putting pressure on terms and pricing. Firms are having to win differently, often at smaller initial deal sizes, while fighting harder for every point of margin.

This is not evidence that the transformation is failing. It is evidence that it is happening in real time, which looks almost identical from the outside and means the opposite.

The minnows

Here is the part I told him I think the industry may still be underestimating, and I have watched this movie twice already.

Right now there are firms you have never heard of doing this work from scratch. Two people and a truck. Onesie twosie engagements. No legacy delivery model to protect, no existing commercial model to unwind, no large organisation to move through. They are not fighting a transition, because they never had the thing you transition away from.

You will not hear about them for another year or two. Then you will hear about all of them at once.

They grow into firms in the low tens of millions. Private equity notices, because private equity notices growth rates. And when a buyer runs the numbers on one of these, the margins do not look like a traditional services business at all. Gross margins up in the sixties where the incumbent model runs somewhere in the forties, with higher net margins layered on top of a materially faster growth rate. Those figures are directional rather than audited, and the direction is the entire engine of what comes next.

So they get bought and scaled. Now they go to market completely differently from you, with a cost structure and an operating model an incumbent finds very hard to match, in front of the same clients.

And then the large incumbent has a decision that is not really a decision. It buys them, or it dies. If it buys them, it becomes the big fish that is genuinely transforming, and it earns the right to become the next holding company or to sell itself to one.

We did this with enterprise software. We did it again with SaaS. We are doing it a third time, and it is going to run faster than either of the first two.

My call, and I will own it: something well short of half the firms you can name today are still standing in five to ten years. Those will transform, get smaller, get leaner, get quicker, buy some people, make it through the trough, and grow again.

The majority get replaced. Not necessarily by each other. By people whose names are not in this conversation yet.

Why the founder came back

There is a detail in this cycle that was not present in the last two, and it is the one I would watch if I were running a large firm.

The founders who already sold and checked out are coming back to do it again.

My friend sees it from the buyer’s side. Established companies want those founders and operators inside the house. A lot of them would rather go build again.

They are not being stubborn. They are being accurate. The reason they needed a large partner in the first place has changed dramatically.

Capital used to be the moat. You needed money for engineers, money for a build, money to carry a sales team long enough to matter. That is what you were buying when you sold most of your company to somebody. AI does not eliminate the capital problem, but it changes its shape entirely. It cuts the capital, the time and the headcount required to get from an idea to a working product.

Their version of it is blunter than mine. Why own twenty percent of the next thing when I could own a hundred percent of it?

That is the supply side of the minnow wave, and it is why I think this one is bigger than the last two. The last cycles were built largely by people doing it for the first time. This one is being built by people who have already been through an exit and know exactly which parts of the old model were load bearing and which parts were theatre.

The forward engineer

He has a term for what I do now that I have heard before. A forward engineer.

I know the business requirements because I lived them. I do not need to gather them. I do not write them down, because it iterates in real time. I do not need to convince anybody or train anybody or sell the project internally. I am recalling thirty years of implementing portals and document management and CRM, and building the thing directly.

The concrete version. I am raising a fund right now, so I built the fundraising platform while doing the fundraising. During the day I am reviewing decks and working on the actual investments. In the evening, when nobody else is working anyway, I iterate on the thing I used that day. By the next morning it exists.

I showed him a piece of it live on the call. There is an assistant sitting over the top of our operating platform now. In the middle of an investment committee meeting I had asked it for a company’s trailing twelve revenue and adjusted EBITDA, and then for the enterprise value multiple on both. It came back with the numbers and with a multiple that was frankly too high, which was itself the useful part.

Then I stopped using it as a search box and started using it as the interface. I typed a plain sentence telling it to update one of our deals to reflect a meeting happening that weekend. It worked out what needed to change, which screen that lived on, made the change, and asked me to confirm it first.

His reaction: you just replaced six apps.

Closer to the truth: six user interfaces I used to have to go and learn. That is the shift people underrate. Not that the machine knows things. That the software stops being a set of screens you navigate and becomes a conversation you have.

The confirmation step is not decoration, and it is the same line I wrote about last week. No AI writes a change to a record without a person confirming it. Speed on the how, humans on the whether.

What it can do that you cannot

I had this out with my own partners a couple of nights before that call. Their position was that there is no way it makes the same kind of decisions a human makes.

They are right, and it is the wrong test.

It cannot make the same decisions you make. Here is what it does that you cannot do. It holds every mathematical formula, every accessible word, and every document in our entire operating system in memory at the same time. Not just tomorrow’s committee materials. Every prior committee, every past document, every result that came out of them, all of it live at once.

Then it recalls and compares all of it at a scale no individual person can. Even with unlimited hours you would not, because retention is not an effort problem. It is a physical limit on what a person is.

A person can imagine one or two scenarios in the time the machine lays out twenty. There is no longer an excuse for not knowing the possibilities.

There is an old line from Napoleon Hill that everybody has heard and almost nobody has taken literally. That which you can conceive and believe, you can achieve. He was describing bending the world toward your own thinking through sheer will.

What changed is the distance between conceiving something and having it exist. That gap used to be funded with capital, headcount and years. Now a great deal of it is just execution the machine will do for you.

My friend put a fair caveat on it. There is still a lot of sweat. It is a different kind of sweat, and it produces exponentially more than the same person could have produced before.

The real barrier is the compensation plan

This is the part of the conversation I would put in front of any board that thinks it has an AI strategy problem.

Agency theory is the idea that people do what is good for them rather than what is good for the company. It is not cynicism. It is just what incentives do. And incentives may be the least discussed barrier to AI transformation there is.

Leaders are being asked to change operating models, organisations, economics, and sometimes the very roles they spent decades building. It is not surprising that reasonable objections emerge. Some of them are legitimate. Some of them are the entirely predictable result of asking people to dismantle the system inside which they succeeded.

Somebody two or three years from retirement or from an exit is not volunteering to be transformed. The objections will all sound operational. Not all of them will be.

Then he raised the broader version of the problem. When senior people are optimising against different incentives, on different time horizons, with different definitions of a good year, enterprise transformation gets exponentially harder.

It produces consequences that look like system failures and are really incentive failures. Cross-selling, collaboration, shared clients, common platforms and enterprise-wide change all get harder the moment the economics reward local optimisation. No technology project fixes that.

So what is an AI transformation actually about? Some of it is technology. Some of it is process. A good deal of it is the operating model. And a far larger part than most companies will admit is aligning incentives with the company you say you are trying to become.

The barrier is not the technology. It is the compensation plan.

Which brings me to the reason the small firms win this. I am looking at companies in the low tens of millions right now that already operate this way and are stuck under the same glass ceiling I was stuck under for years. Getting them to several times that is a distribution and operating problem, and it is solvable.

The thing I cannot manufacture at any price is the part they already have. They do not have to change their culture. Their culture already is this.

You cannot retrofit that. You can start one or you can buy one.

The entry level is where this lands first

One more thing worth saying plainly, because it came up and it is uncomfortable.

The hardest seat in this market right now belongs to inexperienced people working remotely. Not because they are not capable. Because the job now starts with knowing what to tell the machine to do, and then evaluating whether it did a good job. Both of those require judgement, and judgement is the thing you used to acquire by doing the work that has just been automated.

The entire services model was, in large part, an apprenticeship funded by clients paying junior rates. AI is consuming exactly the work that apprenticeship was made of.

Every firm now has to decide whether to train people on work it can no longer bill for in the same way. Most will decide not to. That is not just a talent problem. It is an industry model problem, and it is another reason I think the majority gets replaced rather than repaired.

The other side of it is real too. A friend of mine runs a small advisory practice and now spends most of his working day inside these tools. He automated the entire administrative half of his job. The people part, the part clients actually pay him for, is untouched. He could carry sixty clients where he used to carry six.

That is the same productivity math a large firm is trying to unlock, available to one person with no committee.

The endgame

The holding companies and the large firms at the top of the industry have to go through this transformation or they are finished. They know it. They are also realistic about how much of it they can build organically, which is to say not much. They are going to buy their way there.

Which means the bar for an established firm in the middle of this is different from what most of them assume. You do not have to finish. You have to get far enough down the path that employees, clients, investors and potential buyers can all see the operating model has genuinely changed and that the economics will follow.

Momentum is the product. My guess is a couple of years, give or take, which in the scheme of things is close.

This is not a spectator interest for me. It is the thesis at Bambu Capital, and it needs a better name than the one our industry uses. Everybody in private equity says tech-enabled services, which describes a condition rather than a category and tells you nothing about where the value actually moves.

I have started calling it Services-Tech. Fintech named the moment technology rebuilt how money moves. Proptech did it for buildings, insurtech for underwriting, healthtech for how care gets delivered. Services-Tech is that same event arriving at the businesses that sell expertise, and it sorts them into two groups.

Services businesses that need to become software forward, and services businesses already far enough along to challenge somebody slower. One side transforms. The other side gets transformed. And the same capability decides which is which.

I told him the honest reason I keep in touch, which is that there are not many people I can talk to about the past and the future in the same conversation without having to explain either one. Then I told him the ulterior motive, which is that I do not think he is done.

He is not. And the firms that make it through this will be the ones willing to challenge not just their technology, but their operating model, their commercial model and their incentives, all at the same time.

Because here is the whole thing in one line.

An entire industry priced, packaged and sold the one capability the machine is now best at. It sold how, by the hour, for forty years.

What is left is what, why and if. Nobody has figured out how to bill for that yet. Whoever does owns the next decade of this business.

They sold how.
How is the part that got automated.
What, why and if is the whole business now.

A six part series on what AI is doing to professional services.

One. AI Decides How. Humans Decide What, Why, and If. Governing AI inside a single firm, and where the human veto has to sit.

Two. Nobody Is Buying Hours Anymore. You are here. What happens to an industry that priced and sold the one thing the machine turned out to be best at.

Three. Guess Who Becomes AI Native. Whether this wipes out a generation of young professionals, and what becomes of the pyramid.

Four. The Offerings Transform Too. The second wave, where the platform makes your catalogue unnecessary, and why productising is the wrong answer.

Five. Services-Tech. Naming the category, and the line between the services businesses that get rebuilt and the ones that do not.

Six. Nobody Is Ripping Anything Out. The AI layer over the systems you already own, the migration nobody has named, and why it is the first good news in the series.