The Pyramid That Cannot Fall

I spent most of thirty years inside firms like the ones I am about to describe. Accenture, EY, Oliver Wyman, Capgemini. I didn't study the pyramid from the outside. I built it, staffed it, and sold it. So before I say anything critical, let me be clear about what I think of these firms.

I admire them. The best of them do work that matters enormously. They bring deep expertise into rooms that badly need it, they steady companies through decisions that would frighten most of us, and they have a real hand in innovation and in solving problems that are as political and human as they are commercial. A strong economy needs firms like these.

So what follows is an observation about the model, not a complaint about the people who run it. Every model has a weak point.

Theirs is the pyramid the whole business rests on, and AI has gone straight at it.

What you are really paying for

Start with what a client is actually buying, because it matters for everything after it. You are not really paying for juniors at all. You are paying for judgment: the senior mind that has seen your problem before and knows what to do with it. But that judgment cannot land on your desk out of nowhere. Someone has to gather the facts, run the analysis, test the options, and turn all of it into something a board can act on. That work is real and skilled. For a long time it needed a team, and that team needed weeks. That is where a large part of any bill has always gone. The pyramid grew up as the machinery that carries a client from a blank page to a decision, not as a trick to pad the invoice.

The pyramid does one more thing, and this is where the economics get interesting. The firm bills those junior hours to the client at a heavy multiple of what it pays for them. Stack enough juniors under a partner and the spread on their time becomes the real source of profit. Consultants call it leverage, and it sits close to the heart of how a firm makes money. It also pays for a lot of good: the training, the bench, the years of apprenticeship that turn a graduate into someone worth listening to.

What AI does to that

Now put AI into that machinery. It is best at exactly the legwork the juniors used to carry, the gathering, the analysis, the first draft, the deck. Kate Smaje, who runs technology and AI at McKinsey, was honest about it when she asked whether firms still need “armies of business analysts creating PowerPoints”. Work that took a team of six the better part of a week can now be done by one experienced person in an afternoon. Pass that straight through to the client, and an engagement that cost a million dollars stops costing a million dollars.

Where the value went

So where has all that efficiency gone? The major firms have spent well over 10 billion dollars on AI since 2023, and the money has earned its return. BCG now attributes roughly a quarter of its 2025 revenue to AI services, about 3.6 billion dollars, with AI and technology together above 40% of the business. Accenture reports around 2.7 billion dollars of generative AI revenue in a single year, close to triple the year before. Read those as the firms' own figures rather than gospel, but the direction is not in doubt. The saving didn't reach your invoice. It became their revenue. And the reason has little to do with greed and everything to do with the pyramid.

Why the pyramid cannot fall

Here is the bind. To pass the saving on is to swing an axe at the engine that pays for the partners, the training, and the bench. Put yourself in the managing partner's chair and you'd hesitate too. So the sensible move, for every firm at the same time, is to keep the AI, hold the price, and let the improved margin stay inside the firm.

You can see the strain this puts on the base. Accenture let go of more than 11,000 people in about ninety days, part of an 865 million dollar restructuring, with its chief executive telling analysts it was “exiting on a compressed timeline people where reskilling is not a viable path”. McKinsey is down more than 10% in eighteen months, from about 45,000 people to around 40,000, the deepest cut in its history. Graduate job postings across the Big Four fell 44% in a single year, and in the UK a single graduate consulting seat now draws about 140 applications, up from 86 two years ago. These are careful firms dealing with a real shock, not reckless ones.

The juniors are being taken out of the model. The billing that rested on them cannot come out as fast, and nobody wants to be the first to cut a price, because cutting first looks like admitting the old price was too high. So everyone holds. It is the same standstill you find in any standoff where all sides would gain if one moved and none dares. The result is a strange kind of stability, steady the way a held breath is steady.

What it means if you are buying

Now look at it from your side of the table, because this is the part that matters if you sign the cheque. When the pyramid holds and the AI stays indoors, the price reflects the old machinery more than the work that actually went in. That is not a scandal, just a pricing model built for a time when good thinking was scarce and the only way to sell it was by the hour. That time is passing, and buyers can feel it. The fair question to put to any advisor now is simple.

What happens to my bill when a tool now does in an afternoon what used to take a team of six a week?

If the honest answer is nothing, you have found the pyramid.

The inversion

This is where I should be candid about my own position, and it is not a claim to be better than the firms I came from. A firm with no junior tier has no pyramid to protect. When the whole team is senior, AI is not swapping out a cheap input and keeping the difference, because there is no cheap input. It works on the one thing that was always scarce, the judgment, and lets three experienced people cover ground that used to need thirty.

The effect on price is large. Work a major firm would scope near a million dollars, we can often deliver for a fraction of it, somewhere between a third and a tenth of the cost depending on the engagement. A million-dollar scope coming in at two or three hundred thousand is not unusual. The senior thinking is the same. What falls away is the months of expensive legwork that used to sit between you and it.

That is not a claim to novelty. The tools are the same tools. It is a claim about where the incentives point. In a leverage model, AI is a threat to be contained. In a senior-only model, it is a multiplier with no reason to hide. Same technology. Opposite gravity.

None of this makes the big firm the wrong call. For work that is truly global, truly large, or wrapped in heavy regulation, their scale and their name on the report are the whole point, and worth paying for. The change is narrower and more interesting than the headlines suggest. For work that lives or dies on senior judgment rather than sheer volume, the economics have turned over.

So the biggest firms hold the best AI and the least room to hand you its full value. So much of their profit comes from the very hours the tool now erases, and building AI at their scale, then retraining tens of thousands of people to use it, runs to billions a smaller firm never has to carry. The nimble firm picks up the same tools and passes all of it through, because it was never selling those hours in the first place. The tool is not the advantage. Everyone has the tool. The advantage is having nothing it needs to protect.

Which leaves two questions.

First, once the machine handles the gathering, the drafting, and the first pass, what are you really paying a senior advisor for? And second, what does it actually mean to be AI-augmented, for senior people like us?

That is the subject of part two.