McKinsey’s Newest Analysts Are AI Agents

Every law firm, accounting firm, and consultancy runs on a similar model. Hire a lot of smart young people, bill their hours at a markup, and promote the ones who last. That pyramid has been the most reliable money machine in professional services for a century. Now, artificial general intelligence (AGI) is going after the bottom layer.

What is AGI, you ask? It helps to think about AGI as an employee rather than a tool. Today's AI is a great assistant. It drafts, researches, and summarizes, but it needs supervision, and someone has to check its work. AGI introduces a system that can do most of what a capable person does, for days at a stretch, without anyone looking over its shoulder. In an office, the first person it replaces is not the partner or the rainmaker, but the junior associate at the bottom of the pyramid, the one doing the research and the first drafts that firms have always billed for at a markup.

McKinsey & Company, the privately held consulting firm that advises most of the world's largest companies, is already working this way. When asked how many people McKinsey employs, global managing partner Bob Sternfels told Harvard Business Review's IdeaCast in January that the answer was 60,000: 40,000 humans and 20,000 AI agents. McKinsey's human headcount has fallen from roughly 45,000 to 40,000 since 2023.

Paid for Time or Paid for Access?

In an AGI transition, your pricing model is pretty much your destiny. Paid for time or seats, you lose revenue every time a machine works faster. But if you're being paid for access or outcomes, you can actually gain, because the client's AI still needs your content or your system. We keep coming back to one question for any company in this sector: Does it make more money when its clients need fewer hours?

If the answer is yes, an investor is buying a tier-upgrade cycle, where the company moves each renewing customer onto a higher-priced AI version of the subscription they already hold. RELX (NYSE:RELX) has shown it can do that. A yes also assumes the AI-native rivals sit on top of the incumbents' content rather than replace it, and that's an open question. Harvey, a private legal AI company, says its recurring revenue now tops $400 million, roughly double its March figure, with 80% of top U.S. law firms as customers, according to Bloomberg.

If the answer is no, an investor is buying a repricing race, where the firm has to convert hourly contracts to fixed or outcome-based pricing before clients demand the savings, and before OpenAI's and Anthropic's new deployment arms walk off with the implementation work.

The Fewer-Hours Test

Here are four questions to serve as a starter version of how to evaluate professional services companies for the AGI transition. Four yes answers is Defensible, zero is Vulnerable, and an undisclosed figure counts as a no.

• Does revenue grow faster than headcount?
• Has the company disclosed a metric showing clients pay more per unit of output?
• Is organic growth high single digits or better?
• Does it own content or a system of record that clients have to use?

Legal Research and Law-Firm Software: RELX, Thomson Reuters, Intapp

These are the companies that are paid for access to the case law, the reference material, and the software a law firm cannot practice without, not for anyone's hours. A junior associate's day used to be spent inside Lexis or Westlaw pulling case law and inside the firm's practice-management system logging time and matters. AI now does the pulling, but it does it from the same databases, and the firm still runs on the same system of record. The question is whether the publisher can charge more for the AI version of what it already sells.

RELX, 4 of 4: RELX owns LexisNexis, one of the two databases nearly every U.S. lawyer uses for research. Legal grew 10% underlying in the first half, and three-quarters of renewal value now sits on its Protégé AI tier, a higher-priced version of that same subscription.

Intapp (NASDAQ:INTA), 4 of 4: Intapp makes the software that law and accounting firms run their clients and billing through, which makes it the system any AI agent at the firm has to work inside. Revenue grew 15% while headcount barely moved, from 1,336 to 1,340, with cloud net revenue retention of 123%. That is the small-team leverage an AGI world rewards.

Thomson Reuters (NASDAQ:TRI), 3 of 4: Thomson Reuters owns Westlaw, the other database lawyers research in, plus the tax and accounting content the Big Four cite, and it is bolting the same kind of AI tier onto both. Its core segments grew 10% organically, but it has not yet disclosed whether customers are paying more when they renew.

Consulting: Accenture, With McKinsey as the Benchmark

Ah, consulting. The market-sizing model, the data analysis, the 80-slide deck with one slide that matters, all of it wrapped around a few hours of actual judgment. AI is now excellent at the wrapping. AI builds the model, runs the analysis, and drafts the slides in an afternoon. That leaves the individual who frames the question, directs the agents doing the work, and checks what comes back. It's a real job, but it is a smaller team billing for fewer hours, so the firm has to get paid for the outcome instead of the headcount.

Accenture (NYSE:ACN), 1 of 4: Growth is there, but it has yet to show clients paying more per unit of work, and using AI is not the same as being protected from it.

Outsourcing: ExlService

Outsourcers run the back office for insurers, banks, and health plans, meaning claims processing, finance operations, and data work, and they bill largely per person doing it. AI absorbs that work directly, which is why these firms are racing to reprice before clients ask why they're still paying for seats.

ExlService (NASDAQ:EXLS), 1 of 4: Same story as Accenture. Revenue is growing, but the "AI-led" label on its work has yet to translate into clients paying more per unit.

Research Subscriptions: Gartner

Gartner sells reports and analyst calls that tell IT buyers which vendors to pick. That is synthesis of mostly public information, which is close to what a general-purpose model does for free.

Gartner (NYSE:IT), 0 of 4: Its technology-research client count fell 4.5%.

Recruiting and Staffing: Robert Half

Robert Half places temporary and permanent finance and accounting staff and bills for their hours. The exposure runs two ways. The placed workers do reconciliations, reporting, and bookkeeping, which is work AI is absorbing, and the matching itself could be done by a model.

Robert Half (NYSE:RHI), 0 of 4: Revenue fell about 2.5%. Both it and Gartner point mainly to the macroenvironment rather than AI, and they may be right, but both sell the kind of junior, repeatable work that AI absorbs first.

The Billable Hour Had a Record Year

Here's a twist. The 2026 Am Law 100 results show aggregate revenue up 13%, revenue per lawyer up 8.7%, and profits per partner up 14%. If AI is saving hours, it isn't showing up in clients' bills yet. For now, the gains are going to whoever controls the pricing.

The Bottom Line

The asset best positioned to win is a legal-research renewal book already being resigned at a higher price for its AI tier. RELX has one, and while Thomson Reuters has the Westlaw book, it hasn't yet shown it renewing at a higher price.

RELX reports its third-quarter update on Oct. 22, and investors will be watching legal growth, which has been running near 10%. Thomson Reuters reports Nov. 3, and Wall Street will be looking for any sign that renewal value is migrating to its AI tier at a higher price. That development could move it into the top group.

The Motley Fool has positions in and recommends Accenture Plc, Intapp, and RELX. The Motley Fool recommends ExlService, Gartner, and Thomson Reuters and recommends the following options: long January 2028 $260 calls on Accenture Plc and short January 2028 $280 calls on Accenture Plc. The Motley Fool has a disclosure policy.