Earlier this month, corporate legal operations leaders gathered in Chicago for the CLOC Global Institute. The headline number from their 2026 State of the Industry Report: only 37% of legal departments expect outside counsel spend to grow this year, down from 58% the year before. The work that stopped flowing out is now getting done inside the client.
58%
37%
2025
2026
The shift that’s already happening
Legal operations teams have been building capability quietly. AI has accelerated it. Contract review, first-draft work, regulatory research, certain kinds of due diligence: all of it used to flow automatically to outside firms because in-house teams lacked the bandwidth or the tools to handle it efficiently. That assumption no longer holds for a growing number of companies.
Finance shows the same pattern. FP&A teams that once leaned on advisors for modeling work are running their own scenarios in-house.
This isn’t a competitor eating your lunch. It’s your client deciding they can make lunch themselves.
The distinction matters, because it takes away every move you know. When a competitor improves, you can out-invest them, differentiate, or compete on price. None of the three works on a client. You can’t out-invest your own client, and you can’t differentiate on work they’re already doing. Cutting your price on the work they’re internalizing only speeds up the erosion.
When a competitor gets better, you can compete. When your client gets better, you have to change what you’re selling them.
The tooling reflex
Most firms reach for tooling first. Buy better AI. Run faster. Produce the same work at lower cost so the economics still work. That’s a reasonable short-term answer and a poor long-term strategy.
If your value lived in the production of the work, in drafting, reviewing, analyzing, summarizing, then AI compresses that value whether you use it or your client does. You both get faster. The difference: your client doesn’t pay a billing rate to use their own tools. On your hourly book, that compression is a figure you can run for your firm. You can’t win that race by running harder.
Margin holds in the work the client’s AI can’t do: judgment on the calls where being wrong is expensive, accountability that can’t sit inside the company, and pattern recognition that comes from seeing the same situation across dozens of engagements instead of one.
It’s a concrete change in what your firm puts in front of clients as the thing worth paying for.
Three changes, none comfortable
This is where most conversations stall. Firms agree with the diagnosis and then freeze on the prescription. “Change what you sell” sounds simple until you have to decide what that means for your practice.
Reprice the work AI made faster as a flat fee or retainer. Stop billing hours on it. Clients know the work takes less time now, and billing the old way invites the question of what they’re paying for.
Reposition the work that still belongs with you. The review that surfaces the risky contracts is table stakes now. Sell the judgment call about which ones carry real risk and what to do about them. That’s a different conversation, at a different price.
The work your client can do competently with AI, let them do. You want to be the firm they call when the stakes are high enough that competent isn’t sufficient.
All three mean direct conversations with clients about what the firm is paid for, and partners giving up work they’ve billed for years. They mean building new offerings before the old ones are gone, which is always harder than waiting until the urgency is undeniable.
What still leaves the building
In-house capability doesn’t finish outside firms. It changes what they get called for. The routine work stays in. The higher-stakes work still goes out, and it goes to whoever the client already thinks of as the firm for hard problems.
That reputation gets built long before the hard problem lands, in what your firm decides to sell this year. The client is deciding right now which work they keep and which work they send out. They aren’t asking you which is which.