McKinsey runs 25,000 AI agents. Accenture cut 11,000 people. What the giants are telling you.

When you want to know what a technology will actually do to businesses, do not read the vendor press releases. Watch what the professional services giants do to themselves. They have the earliest visibility into enterprise demand, the biggest incentive to protect their labor model, and the least sentimentality when the math changes.

The math has changed. At CES in January, McKinsey CEO Bob Sternfels described the firm as running on roughly 60,000 "workers": about 40,000 humans and some 25,000 AI agents. Client-facing roles are growing about 25% while non-client-facing roles shrink by a similar amount. Accenture, meanwhile, has been running an $865 million restructuring program, with more than 11,000 roles cut, while posting record bookings and tripling its generative AI revenue. CEO Julie Sweet has been unusually blunt about both halves: the firm is exiting people where reskilling is not viable, and, as she told Fortune last month, the shift amounts to reversing five decades of how the firm works.

Read the structure, not the layoffs

The headlines focus on job cuts. The structural change underneath is more instructive. Consulting firms have always been leverage machines: a small number of partners who own judgment and relationships, sitting on a large pyramid of analysts who produce research, decks, and models. The pyramid was the profit engine. Clients paid partner-level prices for analyst-level work.

AI eats the pyramid from the bottom. Research, synthesis, first-draft everything: this is exactly the work agents do well, and the giants know it because they are deploying it internally at industrial scale. So the shape changes: fewer leverage humans, more agents, and a heavier concentration of value in the people who face clients and make judgment calls. McKinsey's stated goal is every employee enabled by at least one agent. That is not a productivity perk. That is the new pyramid.

The same math is coming for every services business

Private equity has already done this arithmetic. A wave of investors is buying traditional services firms, accounting practices, managed service providers, agencies, and automating 30 to 70% of the delivery work, aiming to exit businesses with software-like margins. Billions of dollars have moved into this "services-as-software" thesis in the past two years. The buyers are not speculating about whether the automation works; they are underwriting deals on it.

If you run a services business of any kind, that is your competitive landscape now. The question is the one the giants already answered: restructure your own leverage model, or wait for a competitor, or a PE-backed roll-up, to do it with your clients.

Three moves translate directly from the giants' playbook to firms a fraction of their size:

• Split judgment work from leverage work, honestly. List what your team actually does in a week. Mark what requires your experience and relationships, and what is structured production: research, drafting, data assembly, formatting, reporting. The second list is your automation roadmap, exactly as we described in the workflows costing your team hours.
• Change what you sell before you change how you staff. The giants' deeper problem is that clients will not keep paying hourly rates for work they know is automated. Firms that move toward fixed-fee, deliverable-based, or outcome-tied pricing capture the automation gain as margin. Firms that keep billing hours hand the gain to their clients and keep the cost structure. The outcome pricing wave is not just a software story.
• Grow the client-facing edge. Note what McKinsey is expanding, not just what it cuts: client-facing capacity, up 25%. In an automated delivery world, the scarce assets are trust, domain judgment, and the ability to sit across from a decision-maker. Investment should follow scarcity.

What it means if you buy consulting

Buyers should update their assumptions too. The economics your vendors run on have changed faster than their rate cards. If a firm is delivering with agent-assisted teams, and the honest ones will tell you, then the price of a deliverable should reflect the new cost of producing it. Ask directly: what share of this engagement is AI-assisted, and how is that reflected in the fee? The answer tells you whether you are buying their new economics or subsidizing their old ones.

It also compresses the case for the biggest brands. As we argued in our honest look at Accenture, the giants remain the right answer for global-scale transformation. But when the delivery layer is increasingly agents plus judgment, smaller senior-heavy firms can field the same tooling with none of the pyramid overhead, and the price difference is no longer buying you anything.

The most honest research report ever published

There is a reason to trust this signal more than any survey or vendor benchmark. The consulting giants have perfect information about what AI does to knowledge work economics, because they are simultaneously the largest sellers of AI transformation and its most aggressive internal adopters. Their restructuring is the most honest research report the industry has ever published.

They are betting their own firms on a specific answer: AI does the leverage work, humans own the judgment, and the pricing model changes to match. Every services business, at every scale, should assume the same answer applies to them, and act while it is still a choice.