The Big Four Are Pricing Themselves to Death

A signal is worth only what it costs to fake. The Big Four just made their signal free. They have priced themselves to death and filed it under efficiency.

Everyone reads the consulting layoffs the same way. McKinsey trimmed around 200 back-office and research roles late last year, with more non-client cuts flagged for the next two years. KPMG is cutting roughly 400 advisory jobs and trimming audit partners. The story writes itself. AI does the analysis, so the analysts go.

The trade press calls this discipline. Margin protection. The firms call it modernisation. It is a confession.

The Conventional Read

Here is the consensus, stated fairly. AI collapsed the cost of analysis. The deck, the model, the diligence pack: a machine now produces in an afternoon what a pyramid of junior analysts billed across three weeks. So you keep the senior judgment, shed the leverage that fed the pyramid, and your margins go up.

Clean. Obvious. The kind of move a CFO approves before lunch. It rests on one assumption nobody at the partner level said out loud: that the analysis was the product, and the analysts were the cost of producing it. That assumption is exactly backwards.

What the Deck Was Actually Doing

The analysis was never the product. It was the receipt.

Think about what a client bought when they paid a seven-figure fee for a strategy engagement. Not the slides. Any MBA can make slides. They bought a costly demonstration that someone serious had committed real, scarce, expensive human hours to their problem before staking a recommendation on it. The 200-page diligence pack was not valuable because anyone read all 200 pages. It was valuable because making it cost something a careless firm could not have afforded to spend.

Biologists have a name for this. In 1975 Amotz Zahavi proposed the handicap principle, and Alan Grafen proved it formally in 1990: a signal is honest in equilibrium if, and only if, its cost is prohibitive for a faker to bear. The peacock's tail is not pretty by accident. It is expensive on purpose. Strip the cost out and mimics flood in until the signal means nothing.

The deck, the model, the diligence pack were all costly signals. Their honesty was load-bearing on their price. Now automate them.

The Trap, Stated as a Theorem

By Zahavi's own logic, a signal that becomes cheap to produce stops signalling anything. The cost was not a flaw in the work the firms have just optimised away. The cost was the work. So the Big Four did not cut cost. They cut credibility.

We call this The Unhedgeable Signal, and the logic chain fits on one line. The only thing a buyer can trust is a cost a faker cannot pay. The only cost nobody can offload onto someone else is the bearing of true uncertainty, the judgment that cannot be insured or hedged. And that uncertainty-bearing is precisely the part of the work that does not parallelise, the irreducibly serial fraction that no amount of compute can split apart.

Three fields describe one object: value that exists only because a cost cannot be faked, hedged, or automated. AI is a machine for making costs hedgeable. Point it at advisory work and it drives the price of every analysable, parallelisable task toward zero. By the theorem, that does not lower the price of the signal. It raises it, by destroying everything cheap around it. The premium does not disappear. It migrates into the one fraction AI cannot touch: the willingness to make the call, on the record, and own being wrong.

The Part They Kept Is the Part They Cannot Do Anymore

Here is the cruelty in the org chart. The Big Four kept the senior partners, the brand, the overhead. They shed the juniors. But the juniors were never just cheap labour producing now-automated decks. They were the apprenticeship. They were the people slowly, expensively learning to bear the uncertainty that becomes the only billable thing left.

You have removed the bottom of the funnel that manufactures the judgment at the top. You are keeping the expensive overhead while firing the people who were learning to carry the one thing that still commands a premium.

That is not a leaner pyramid. That is a pyramid eating its own foundation and calling the hollow space a margin.

Christensen Is Already in the Room

This is disruption from the bottom, textbook Clayton Christensen. The incumbents overserve. They sell brand-heavy, high-touch engagements priced for the largest balance sheets. AI just made the analytical floor of that service free for everyone. So the underserved buyers, the scaleups and mid-market firms who never needed the theatre, can now assemble a fractional expert, an AI workflow, and a boutique specialist into something good enough, then better.

The pricing model flips with it. When analysis is free, you cannot bill for hours. Hours were always a proxy for the cost-signal anyway. What survives is outcome-based pricing: charge against the result, against the willingness to put your name on the bet. Not the time spent making the receipt. The disruptors start with the boring, repeatable jobs the incumbents priced out of reach, then climb. The majors are accelerating it by hand, removing the cost-signal from their own product and the apprenticeship from their own bench.

The Strongest Case for the Other Side

Concede the steelman in one line: AI genuinely raises margins, and buyers do not romanticise process, they want the outcome. The cost-signal was always a tax on the client, and AI is repealing it.

Now follow it through. The whole problem with a high-stakes advisory call is that you cannot verify the outcome at the moment of purchase. That is what uncertainty means. If you could check the answer up front, you would not buy trust at all. You would buy the answer. The cost-signal exists precisely because the result is unknowable when the cheque is signed. Remove the signal and you have not given the client a cheaper result. You have given them an unverifiable claim with nothing expensive standing behind it.

The buyer does only care about the result. That is exactly why they pay for the one thing that credibly predicts it: a firm willing to bear the cost of being wrong. Automate that away and you have not won the buyer. You have become the free tool they could have used themselves.

The Close

You cannot automate the part that costs you something. That part was never the analysis. It was the willingness to be wrong on the record. As we argue in a companion series on professional services becoming a financeable asset, the firms that price conviction honestly are the ones capital learns to trust.

Price the analysis at zero. Charge for the conviction. They have done the first half and forgotten there was a second.