Your firm’s best AI workflow may already exist—and no one else knows about it

Nearly every law firm already has an AI adoption curve hiding inside it. Some people have not begun. Others are testing ChatGPT or Claude, sometimes through personal accounts the firm cannot see. And one lawyer, paralegal, or technically curious staff member is often far ahead—quietly using AI to improve a recurring part of the work.

That unevenness is not evidence that the firm lacks interest. It is evidence that experimentation has outrun leadership. Useful techniques remain personal, risks remain inconsistent, and the firm receives little institutional value from what its strongest users have learned.

A power user is not an operating model

A strong individual workflow can create real value, but it is fragile. It may depend on one person’s prompt history, personal account, undocumented judgment, or willingness to help colleagues between matters. If that person leaves, gets busy, or misunderstands a risk, the firm has no durable capability to fall back on.

The goal is not to suppress experimentation. It is to create a path by which good experiments become visible, reviewable, teachable, and safe enough for others to use.

Start with the work, not the tool

AI leadership begins by listening to the people closest to the workflow. What do paralegals repeatedly assemble, compare, extract, or chase? Where do associates lose time moving between sources? Which first drafts, chronologies, reviews, or client updates already benefit from AI? Who has found a better method, and what judgment makes it work?

The leader’s job is to map the inputs, decisions, review points, source material, and failure modes around that method. A clever prompt is not yet a firm workflow. It becomes one only when the firm can explain when to use it, which tools and information are permitted, how output is checked, and who owns the result.

Turn local learning into firm capability

Once a workflow proves useful, capture more than the prompt. Preserve the technique, approved tools, examples, source requirements, review standard, and known limits. Train the people who perform that work, identify internal champions, and give them a place to share improvements without creating a new shadow system.

Not every lawyer needs to use AI identically. Practice groups differ, matters differ, and professional judgment remains personal. But the firm should have a shared baseline, reusable patterns, and a governed path for turning individual learning into collective advantage.

That is the leadership gap

Tools do not discover the best work inside a firm. A policy does not spread it. A training session cannot sustain it by itself. Someone must connect the people experimenting, the workflows worth improving, the risks the firm must control, and the operating habits that make learning repeatable.

The firms that pull ahead will not simply have more AI users. They will get better at learning from their best ones.