$100 Billion of Consulting Value Evaporated. The Market Wasn't Wrong.

A few years ago, at Adaptavist, we walked away from a six-figure contract with a very large investment firm.

They wanted us to start building. In pre-sales we kept trying to get at the why – what outcome are we actually after, what does success look like – and the answers weren't just vague, some of them contradicted each other. We proposed a discovery engagement first, to help them work out what they actually wanted. They declined. They wanted to start building and sort it out along the way.

So I told them, roughly: you don't know what you want, you don't want to do the work to figure it out, and you're not accepting our help to figure it out. If we take this as it stands, you won't be happy with the outcome, we won't be happy, and it'll damage the relationship. The better move is for us to politely decline.

We fully expected that to cost us the contract. It wasn't a negotiating tactic – we said it because it was true. They loved the answer, and they signed up for the discovery engagement instead.

I've been thinking about that conversation while reading what the equity markets have done to the consulting sector this year.

What actually got repriced

Reuters Breakingviews reported in late July that Accenture and Cognizant have lost more than half their market values – a combined $100 billion – over the past two years, as investors price in a future where AI commoditizes such expertise. Everyone from McKinsey to Grant Thornton, the piece notes, now faces the challenge of proving the value of their strategic counsel by implementing it in their own firms first.

This isn't a sudden repricing. Breakingviews was already making the case in October 2025, when shares of the largest publicly traded firms had fallen as much as 30% over two years while the S&P 500 rose 50%, in a sector worth roughly a trillion dollars. And Accenture's near-20% single-day fall in June 2026 was the worst day in its history as a public company, on softer guidance and a decline in new bookings.

The near-universal read of this is that AI is killing consulting. I think the market is being considerably more specific than that, and the specificity is the whole story.

Labor is what's on sale

The economics of large-scale professional services run on leverage: a partner sells the engagement, and the work gets done by a pyramid of people billed out at multiples of what they cost. Margin comes from the spread between billing rate and salary, multiplied by headcount and utilization. Scale the pyramid, scale the profit. It's an elegant model and it built some of the most durable firms in the world.

Now automate the bottom of the pyramid.

The research synthesis, the document review, the first-pass analysis, the deck assembly – the work that junior people did, learned from, and were billed out for – is exactly the category AI handles competently and cheaply. So the revenue tied to that work compresses, the margin structure built on it compresses harder, and an investor looking at a business whose earnings depend on billing hours for automatable work marks it down. That's not a verdict on whether advice is valuable. It's a verdict on a specific mechanism for converting advice into revenue.

Which is why I'd resist the "AI is killing consulting" framing even though the numbers are real. The market didn't decide that judgment is worthless. It decided that judgment sold by the headcount-hour has a cost structure that no longer holds, and those are very different conclusions with very different implications depending on which business you're in.

The thing that didn't get marked down

Go back to that investment firm.

What we sold them in that room wasn't analysis. It was a no. Specifically: a person who had seen enough of these engagements to recognize that the thing they were asking for would waste their money, and who was willing to say so at the cost of the contract.

There is no version of that conversation that a system optimized to be helpful produces. The reasoning isn't especially hard, and you could probably get a model to lay out the risks of an underspecified engagement. What the model doesn't have is anything at stake. It isn't the one who has to sit across from that client in eighteen months when the thing they built turns out to be wrong. It has no relationship to damage, no reputation that absorbs the failure, and no ability to be wrong in a way that costs it anything.

An agreeable, fast, plausible yes is now essentially free. Which means it's worth roughly what it costs.

That's the actual asymmetry the write-down exposes, and it cuts in an unexpected direction. What got cheap is the production of analysis. What stayed expensive – and got scarcer relative to the flood of cheap output – is somebody with the standing and the exposure to tell a client the uncomfortable thing, and to be accountable for having said it.

What this means if you work alone

If you're a senior independent consultant or you run a small advisory firm, the temptation right now is to read this news as a storm you need to survive.

I'd read it almost the opposite way. You were never in the business that got repriced.

You don't have a pyramid. There's no bench of juniors whose billable hours are being automated out from under you, no utilization model to defend, no leverage ratio that stops working when first-draft analysis becomes free. The thing you sell is the thing that stayed expensive, and you've been selling it without the pyramid the whole time.

The practical shift is smaller than the headlines suggest. Cheap analysis raises the floor for everyone, including your clients, who can now generate a serviceable market overview themselves in an afternoon. So the deliverable that was a competent summary of the landscape is worth much less than it was, and it should be – it's the part a machine now does well. What your client cannot generate is a read on which of the three plausible options is actually right for their situation from someone who has watched that decision play out before and whose name is attached to the recommendation.

Let the machine do the reading. The judgment about what matters was always the product.

The uncomfortable part

The same forces that compress the pyramid also raise the bar on what counts as judgment. If your work has quietly been synthesis with a light seasoning of opinion – the reading, the summary, a recommendation that follows fairly directly from the summary – then a good chunk of what you do is now reproducible at low cost, sole practitioner or not. The protection isn't structural. It's whether your recommendation contains something the analysis doesn't already imply.

That's a real question, and the answer for most of us is that some of our work passes it and some doesn't.

The market repriced advice that arrives as labor. It hasn't yet found a way to reprice someone who will tell a client the truth and put their name on it, because that has never been what the pyramid was selling.