Paving the Path to Operational AI: What the Firms Making Progress are Doing Differently

Firms have proven they can adopt AI. The challenge now is redesigning workflows, breaking down silos, and turning productivity gains into business outcomes.

By Mike Raposa

At ILTACON this year, Williams Lea hosted a private roundtable for innovation and AI leaders from a group of AmLaw firms. Ari Kaplan moderated. Two long tables, one evening, no slides and no presentations. The only thing we asked for was candor, and the room gave it.

What has stayed with me is how the conversation moved. It opened on real enthusiasm and ended somewhere much more honest. Both of those moments deserve attention, because the distance between them is the work the profession has in front of it.

Adoption is real, and firms have backed it

The people in that room accomplished something the legal industry spent two decades failing to do. Participants described adoption at a speed none of them had seen in their careers. One described a partnership running at roughly ninety percent active use on a single tool. Another talked about lawyers adopting by the tens of thousands on a compressed timeline. For anyone who remembers how long it took this profession to accept email, that is remarkable.

More to the point, firms have put structure behind it. Several participants described incentive systems built specifically for this work: AI credit hours that count toward bonus, innovation pools measured in hundreds of hours, and committees that review a proposed workflow, validate that it actually executes, and confirm it is reusable before any credit is awarded. One firm opened its program on September 1 and had roughly fifty associates approach with ideas before the month began. Staff and paralegals are eligible in both of the programs described that night.

This tracks with what is happening across industries. Nearly nine in ten organizations now use AI regularly in at least one business function, and 44 percent report scaling it across the enterprise, up from 38 percent a year ago (McKinsey, State of AI 2026). Legal has funded its share. Technology and knowledge management spending grew 11.6 percent overall and 8.7 percent per lawyer in the second quarter of 2026, part of a multi-year investment trend (Thomson Reuters, Q2 2026 Law Firm Financial Index).

The adoption argument is settled. The question worth asking now is where do firms go from here. 

Where firms stall

Here is the finding that should reframe how firms think about their AI investment. Eighty percent of respondents in McKinsey’s most recent global survey say AI has improved their individual productivity. Thirty-seven percent say it has contributed to their organisation’s EBIT, a figure essentially unchanged from the year before, and the share of AI high performers held flat at about six percent. McKinsey’s own summary is that conviction in AI is growing faster than the financial returns organizations can attribute to it (McKinsey, State of AI 2026).

That gap has one dominant explanation, and the same survey names it. Nearly three-quarters of high performers report fundamentally redesigning workflows because of their AI use, up from 55 percent the prior year. Among all other respondents, the figure is one quarter (McKinsey, State of AI 2026).

Capability laid over an unchanged process gives you faster individuals and the same business outcome. The technology works. The work around it kept its original shape. Legal shows this pattern clearly. Thomson Reuters found that at many firms AI has changed which tools lawyers use while leaving intact how matters are run, staffed and delivered. Half of law firm professionals say their firm’s AI strategy is invisible in their day-to-day work (Thomson Reuters, Future of Professionals Report 2026).

What the firms making progress do differently

This is where the roundtable earned its evening, because the answer came from the room.

A participant observed that firms have always operated vertically by department. Marketing here, finance there, records and IT somewhere else. Each function owns its own output, and that arrangement works until the deliverable is an outcome crossing all of them. AI-enabled workflows are exactly that kind of deliverable. They need the data to be right, the output to be credible, the format to hold up and the knowledge base behind it to be current.

Then the senior AI and knowledge management leaders in the room said something I have not heard said publicly. They cannot be the ones building every workflow the firm will ever need. The people this industry has made accountable for redesigning the work told a table of their peers that the job requires business services, marketing, finance and knowledge professionals present from the beginning.

That is the clearest pattern I took away from the night. The firms making the most progress are designing workflows with the people closest to the process in the room from the start.

Governance follows the same logic. Several participants drew a line between governance for the practice of law and governance for the business of law, where a firm owns its data the way any enterprise does. Treating those as a single problem slows both of them down.

The hardest part is prioritization. A participant who ran enterprise transformation before joining a firm put it directly: transformation works when you prioritize, and a partnership with hundreds of owners finds that structurally difficult. His own default answer, he said, is maybe. He never says no. Choosing what to build first, sequencing what follows, and holding the authority to stop something are ordinary operating disciplines, and they are now on the critical path.

The research points the same way. High performers are twice as likely as their peers to say senior leaders demonstrate commitment to AI initiatives and to report defined processes for measuring impact (McKinsey, State of AI 2026). Those are governance capabilities. And McKinsey’s work on law firms argues that preeminence increasingly depends on platform infrastructure, with institutional knowledge moved out of document management systems and individual practitioners into structured, reusable assets (McKinsey, Shaping the Law Firm of the Future). That is an enterprise build, which is precisely why it takes the enterprise to build it.

The layer the conversation keeps missing

Late in the evening, after two hours spent almost entirely on the practice of law, a participant asked whether any of this effort was reaching the operational layer. The administrative workflows. Document processing. Billing.  The services that connect people, systems, offices and business functions and underpin the practice. 

It was the right question, and it came last. That sequencing tells you something about where the profession’s attention has been.

Clients are already measuring the difference. Seventy-seven percent say AI-enabled quality improvement from their firms is very important or essential. Between three and six percent believe most or all of the firms they work with are delivering it (Thomson Reuters, Future of Professionals Report 2026). A gap that wide closes from one direction or another.

We ended the night going around the table for takeaways, and the group landed on the fact that nobody has fully figured this out yet. I find that encouraging. Firms have proven they can adopt. What sits in front of them now is operational: redesigning how work moves, deciding what to build first, and keeping the whole firm in the room while it happens.

That is the path to operational AI, and it is worth paving properly.

– Mike Raposa is Chief Revenue Officer at Williams Lea.