The analysis got cheaper. The judgment didn’t.

Every boutique firm hears the same question early in a first conversation: how big are you? For most of consulting's history, it was a fair question. AI is changing what the answer means.

In January, Bob Sternfels, who leads McKinsey as its global managing partner, gave an unusual answer for his own firm. Speaking on Harvard Business Review's IdeaCast and at the Consumer Electronics Show, he described a workforce of roughly 40,000 people and about 25,000 AI agents, up from a few thousand agents eighteen months earlier. He expects the two numbers to be equal by the end of this year.

It is a striking way to answer the question. It also suggests the question has changed.

Many organizations still judge a consulting firm by its size, and still pay for proposals built on large teams of junior staff. Increasingly, that means paying for work AI now does in hours, while the experience that decides whether a program lands is spread thin. To see why, it helps to look at how headcount became the measure in the first place.

Consulting began by selling experience

In its early decades, consulting sold what experienced people knew. Arthur D. Little, a Boston chemist, founded what is widely called the first management consulting firm in 1886, selling technical research and analysis. James O. McKinsey, a former accounting professor at the University of Chicago, founded his firm in 1926, and his best-known contribution was showing executives how to use the budget as a tool for managing.

Clients hired these firms for judgment they could not find inside their own walls. There was only so much of it to go around.

Then headcount became the business model

The industry's defining shift came later. Under Marvin Bower, who shaped McKinsey for decades after its founder's death, the firm became known as the first to staff its projects with newly graduated MBAs rather than experienced hires from industry. It was a powerful model. A few senior partners could win the work and stand behind it, while teams of capable young analysts did the research and built the analysis underneath them.

David Maister, a former Harvard Business School professor who spent his career studying professional firms, called the ratio of junior staff to partners leverage. It became the engine of the industry's economics. The more people a firm could put beneath each partner, the more work it could take on, and the more it could earn.

So for decades, “How big are you?” was a reasonable question to ask a consulting firm. Size meant capacity. More people meant more analysis, more documents, and more work done faster.

What AI absorbed

The work at the base of that pyramid is exactly the work AI does best. McKinsey says AI saved it about 1.5 million hours of search and synthesis in a single year. Across the largest consulting firms, revenue kept growing through 2025 while headcount did not. McKinsey and other top firms have frozen starting salaries for a third straight year, finding that AI lets them get more from fewer junior staff.

It would be easy to read that as the end of the pyramid. It isn't, quite. McKinsey still plans to grow its North American workforce by 12 percent in 2026. What is changing is where the people go. By the firm's own account, its client-facing roles have grown by about a quarter while the rest of the organization is streamlined. Meanwhile, a new kind of firm is appearing, as former Big Four partners launch boutiques built around AI from the start.

In other words, one of the most influential firms in the industry is using AI to move its people toward its clients.

What we see at SIC

We see the same shift at our own scale. At SIC, AI has taken over much of the work that used to fill the hours between client meetings. Meetings are captured as they happen. Decisions are documented and can be found quickly months later, when someone asks why the project went one way and not another. Data analysis and timeline organization move faster, and once the tools learn what we need, charts, tables, graphs and presentations take a fraction of the time they once did.

The difference shows up in where our time goes. More of it is spent in front of clients. Changes that used to mean taking the work away and bringing a revised version to the next meeting now happen in the room, while the people who need to see them are still there.

What AI cannot carry

This matters most in the kind of work SIC does. Implementations rarely fail for lack of analysis. They fail when decisions wait, when the people who have to change how they work are not brought along, or when no one owns the moment a go-live is at risk.

AI can help draft the plan, check the data and organize the timeline. It cannot make a county's decisions for it. It cannot sit with a department head who is worried about what a new system means for her staff, or tell a steering committee the truth about a date. It cannot stand in a go-live and decide whether to proceed.

Asked what AI cannot replicate, Sternfels named three things: setting aspirations, human judgment, and genuine creativity. In an implementation, judgment is most of the job.

So does headcount still matter?

It does. It just measures something different. Some programs genuinely need more hands, and providing the right ones is part of what we do. But headcount used to stand in for analysis capacity, and that capacity is now cheap. What remains scarce is experience: people who have watched a program slip before and know what to do when it starts to.

That changes the useful question for anyone hiring a firm. Not how many people will be on the team, but who they are, what they have done before, and how much of the work they will do themselves. A proposal built around a large team of junior staff deserves a closer look now that the analysis is the part that got cheaper.

How big are you?

When someone asks us that question, our answer has never been a number. We staff by selection rather than availability, and every engagement is led by an experienced professional chosen for that program, with SIC's wider network behind them when the work needs more.

AI has not changed that answer. It has made it a stronger one.