Nov 11, 2026

Law Firms Have Adopted AI. The Returns Depend on How They Run the Business

By Mike Raposa, Chief Revenue Officer, Williams Lea

Here is the uncomfortable truth about AI in law firms: adoption was the easy part. Firms did in two years what this profession failed to do with almost every technology before it. Ninety-four percent of firms in this year’s ILTA Technology Survey are now engaged with generative AI, up 14 points in a single year, and technology and knowledge management spending rose 11.6 percent in the second quarter of 2026, according to the Thomson Reuters Law Firm Financial Index. Anyone who remembers how long it took lawyers to accept email should find that remarkable. I do.

But adoption was never the goal. Return is. And in return, the evidence is starting to come in—and it should get every managing partner’s attention.

Faster people, same business

McKinsey asked more than 1,700 executives what AI has done for them. Eighty percent said it made them more productive. Only 37 percent could point to any effect on earnings, and that number has barely moved in a year.

Read those two numbers together. Organizations everywhere have made their people faster and left their businesses the same. The same survey explains why. Nearly three-quarters of the companies getting real value redesigned their workflows around AI. Among everyone else, one in four did. Deloitte found that 84 percent of companies have yet to redesign jobs around the technology at all. Put the most capable tool ever built into a process designed decades ago and you get a faster version of decades ago.

Law has less room for error than anyone

That gap matters more in law than in any industry I know, because growth has been running on price. Citi Hildebrandt’s 2026 Client Advisory shows revenue up 11.3 percent through the first nine months of 2025, on rates up 9.6 percent and demand up 1.9 percent. That is a business charging more for roughly the same amount of work, while support staff costs climb more than 6 percent a year, according to the Thomson Reuters Institute’s 2026 State of the US Legal Market report.

Now add a technology whose entire promise is doing the same work in less time. Every large firm Citi surveyed plans to spend more on generative AI. Citi asks how those firms will recover the cost. I have yet to hear a convincing answer.

Clients have not forced the issue. Most general counsel still are not requiring their firms to use AI. But their own legal departments have been ahead of law firms on AI since 2022, and they can see exactly what it does to routine work. They are going to ask what their firms have done with it, and the firms with a good answer will be the ones that changed how the work gets done.

What I heard in the room

Late this summer I sat down with innovation and AI leaders from a group of AmLaw firms for an evening of unusually candid conversation. The adoption stories were impressive. One partnership was running at roughly 90 percent active use of a single tool. Firms had built AI credit hours that count toward bonus, and review committees that award no credit until a workflow actually runs and can be reused. One firm opened its incentive program on September 1 and had about 50 associates lined up with ideas before the month began.

Then the AI and knowledge management leaders said something I had never heard said out loud. They cannot build every workflow their firms need, and they should not be expected to.

Think about what that means. The people this industry has made accountable for AI are telling us the work crosses every department in the firm: marketing, finance, records, IT and knowledge. Law firms have always run side-loaded by department, organized in a way that keeps those people apart. An AI workflow needs clean data, credible output, a format that holds up and a knowledge base that stays current. No single function owns all of that. So the workflow has to be designed by all of them, together, from the first day.

Here is the irony. Firms already employ these people, and pay well for them. Large firms have spent five years adding business development, technology, pricing, billing and finance professionals, and Citi counts that hiring among the major drivers of expense growth. Too often, in my experience, those professionals are handed the finished tool and asked to support it. That is an expensive way to waste expertise. Deloitte’s research across industries points the same direction: the organizations seeing value bring IT, legal, compliance and business leaders together early, “so that scale does not outpace control.”

Maybe is the most expensive answer in law

The hardest part is choosing. A partnership with hundreds of owners is built to keep options open, and the easiest answer to any new idea is maybe. With AI, maybe has a price. Deloitte calls it the proof-of-concept trap: projects scoped at three months that stretch to 18, and only a quarter of organizations moving even 40 percent of their experiments into production. Any firm running a dozen pilots at once should recognize itself in that.

The organizations that escape the trap are led differently. McKinsey’s high performers are twice as likely to have senior leaders who champion AI over time and defined processes for measuring what it returns. Thomson Reuters found that law firms with a visible AI strategy are 3.9 times as likely to see a return at all. Deciding what to build first, what follows and what to stop is ordinary operating discipline. In a partnership it has always been hard. Now it is the critical path.

Start where the money moves

If I could move one thing up every firm’s list, it would be the operational layer: the billing, collections, document processing and back-office services running underneath the practice. That is where redesign shows up directly in margin, and it is where the conversation has spent the least time. ILTA says it plainly: the biggest factor in the year ahead “won’t be GenAI itself; it will be the business process discipline, data strategy, and systems integration required to use it well.” Only 60 percent of firms run time and billing in the cloud, and the average collection cycle has stretched from 115 days to 134 in five years, according to Citi Hildebrandt. Those are operating problems, and AI can finally help solve them once someone redesigns the work.

The stakes go beyond margin. Citi expects leverage to shift from a pyramid toward a cylinder, and 36 percent of large firms expect generative AI to change their staffing models within two years, nearly double last year’s share. Firms that redesign the work on purpose will choose their new shape. The rest will discover it.

We ended that evening by going around the table, and the group agreed on one thing: nobody has this figured out yet. I find that encouraging. It means the window is still open. Law firms have proven they can adopt. Now they have to prove they can run differently, by redesigning how work moves, deciding what to build first and bringing the whole firm into the room while it happens.