AI Consulting is Eating Itself

There is a contradiction at the center of the current AI consulting boom that the industry has not been particularly eager to examine.

Firms are billing clients at record rates to navigate an AI transformation — and using that same AI to eliminate the analysts, associates, and junior staff who used to do the work those bills are paying for. The pitch and the backstage operation are running in opposite directions, and everyone involved seems to have agreed, tacitly, not to bring it up.

This isn't an indictment of the industry. Consulting isn't going anywhere, and the firms driving this shift are not uniquely cynical organizations. But something real is happening to the model — something that hasn't been fully named yet — and it matters for anyone trying to figure out what they're buying when they hire a firm to help them think through AI.

The Pattern

The numbers from the last few years tell a consistent story. McKinsey shed roughly 5,000 employees between 2022 and mid-2025, with another significant round announced not long after. Accenture eliminated 11,000 roles and pledged billions in AI investment in the same press release — both stated plainly, as though the tension between them didn't warrant comment. Junior hiring across the industry quietly contracted as AI absorbed the analytical entry-level work that once filled those seats.

The fact worth sitting with is this: roughly 150 former McKinsey, Bain, and BCG consultants were contracted to train the AI models automating the tasks they had themselves once performed. They weren't passive victims — they took the work willingly, and in their position most people would have. But the fact that the work existed, that there was a functioning market for it, says something about the direction the industry was already heading.

Global consulting revenues still grew around 5.5% in 2025. You could read that charitably — leaner firms, same output, better margins, more projects. The less charitable reading is that clients aren't seeing much of that efficiency. Both things are probably true to some degree. The margin between them is where the honest conversation about this industry should be happening.

What the Model Was Built On

Large consulting firms were designed around a specific kind of leverage: senior judgment, distributed through a pyramid of billable bodies. Partners brought relationships and pattern recognition. Everyone below them did the research, the modeling, the slide-building — work that could be packaged, billed, and scaled. The pyramid wasn't incidental to the business. It was the business.

AI collapses the bottom of that pyramid. The research gets done faster and cheaper. The first-draft recommendations write themselves. The benchmarking that used to take a week of associate hours takes an afternoon. This is genuinely good for productivity. It is genuinely bad for a billing model premised on those hours existing.

Deloitte announced in early 2026 that it's scrapping traditional job titles as part of an AI-driven restructuring. The pyramid isn't just shrinking. The shape itself is changing, and no one inside these organizations seems entirely certain what it's supposed to look like on the other side.

What Clients Are Actually Experiencing

The clients we work with don't arrive energized by possibility. They arrive with a specific kind of pressure — the sense that something important is happening around them and they haven't figured out their place in it yet. That feeling makes people reach for things that look authoritative: recognized names, thick proposals, methodologies with acronyms.

What they often get is advice calibrated for a company that resembles theirs — not the specific, operationally complicated organization they're running. The recommendations are defensible and sound good to board members. The team pivots in, oftentimes, superficial ways. Six months later the organization is more or less where it started, except now it has a strategy deck.

This isn't a talent problem. It's a scale problem. You cannot build a multi-billion dollar practice on engagements that require genuine, sustained proximity to a single client's reality. So the industry built frameworks instead — and for a long time, clients didn't have a clear enough view of the alternative to know what they were missing.

Where the Value Actually Lives

If AI can now approximate the research, the modeling, the first-draft thinking — and it increasingly can — then what remains is something harder to systematize: the judgment to know which question is actually worth asking, the fluency to operate inside a specific organization's culture and constraints, and the accountability that comes from staying in the room when things get complicated.

Those things don't scale easily. They're also the things clients actually needed in the first place.

There's a version of this work that looks less like a consulting engagement and more like sitting with the people who actually run the operation, following a real process from start to finish and not leaving until something has materially changed. That kind of proximity doesn't produce a better deck. It produces a different outcome entirely.

Where This Goes

The large firms will survive. They'll do what incumbent organizations do when they encounter a threat they can't build their way out of — they'll acquire it. In the meantime, there's a window. Clients are getting more sophisticated about what they're actually buying. The question is no longer who has access to AI — it's who has genuine fluency with it, and the willingness to stay in the problem long enough for it to change something.