Drafting has already changed hands

Law Insider's 2026 study of AI in transactional work carries a statistic that has had less attention than it deserves (lawinsider.com/reports/legal-ai-study-2026). Among 534 surveyed legal professionals, all of them AI adopters and many of them power users, 73% use AI for drafting clauses and contract language. It is the single most common use of AI in contract work, ahead of review, ahead of summarisation, ahead of everything.

Brief disclosure: I am CPO at SimpleDocs, whose product appears in this study, and Law Insider is our close collaborator. This piece is not about tools, but you should know where I sit.

Now put the drafting number next to the frequency data. 86% of these lawyers use AI on contracts at least weekly and 41% multiple times a day. The profession's most identity-defining task, the one juniors are trained on, the one that fills the first five years of a transactional career, is now routinely initiated by a machine. Not in a pilot programme. In the daily production of contracts, across 75 countries, at every firm size from solo practice upward.

And here is the detail that turns this from an efficiency story into something else entirely. These same lawyers do not trust the output. No tool in the study persuades even one in five of its own users to say they are very confident in what it produces. The most trusted tool manages 18%.

A profession that drafts by machine and believes the machine 18% of the time at best has not automated drafting. It has converted drafting into something new: a verification discipline. The lawyer's contribution has moved from producing the language to interrogating it. That shift has already happened, invisibly, one lawyer at a time, and almost nothing about how the profession trains, prices, or organises itself has moved to reflect it.

Verification is the actual job now

Consider what a transactional lawyer using these tools actually does in a day, given the numbers above. The machine produces a clause in seconds. The lawyer then does the real work: is this consistent with what we agreed last time, does it match positions we can defend, does it fit what the counterparty will accept, is the confident phrasing hiding an error. Every one of those questions is a verification question, and every one draws on knowledge the machine was not given.

The study shows lawyers know exactly where that boundary sits. AI usage collapses precisely at the tasks where verification requires institutional knowledge: comparing against internal precedent (23%) and applying playbooks and fallbacks (16%). And when 344 lawyers described what they most wanted from their tools, the largest cluster was verification and sources. Read as a labour signal rather than a feature request, that is a workforce describing its new bottleneck. The scarce activity is no longer generation. It is the checking.

There is a version of this observation that sounds like diminishment, as if lawyers have been demoted to proofreaders of machine output. I think the opposite is true, and the trust data is the evidence. What survives contact with a fluent machine is exactly the part of legal work that was always the point: judgment about acceptability, memory of what this organization has done, sense of what the market bears. The machine has stripped away the typing and exposed the judgment underneath. What should worry the profession is not the shift itself but how unprepared its structures are for it.

The training pyramid was built on the work that just left

The transactional career has a shape everyone in it recognises. Juniors draft. They draft badly, then adequately, then well, and somewhere in those thousands of hours of producing language under correction they absorb the judgment that makes a senior lawyer valuable. The drafting was never really the product. It was the apprenticeship.

The study describes the base of that pyramid being handed to machines by the very people who climbed it. 73% machine-assisted drafting, concentrated among engaged early adopters, means the formative work is disappearing from the junior years first and fastest. The judgment that verification requires, the thing the whole system now runs on, was traditionally a by-product of doing the drafting yourself. We are removing the by-product's source while betting the profession on the by-product.

Nothing in legal education has responded. Law schools still teach as if graduates will spend a decade producing language before being trusted with judgment. Firms still price juniors as drafting labour. The billable hour still assumes the time-consuming part is the writing. Meanwhile the actual daily work, on the evidence of this study, has become the interrogation of machine output against organizational and market knowledge that no junior possesses and no current training path builds deliberately.

The institutions that produce lawyers are training people for the job that just ended.

Judgment has to be taught on purpose now

If verification is the job, the profession's central question changes. It is no longer how to make lawyers faster at producing documents. It is how to build the judgment that verification runs on, deliberately, now that the accidental apprenticeship of drafting is going away.

Part of the answer is institutional and falls on legal teams. The knowledge a lawyer verifies against, prior positions, real standards, market reality, currently lives in signed PDFs and senior memories. Teams that make it explicit give their juniors something to learn from and their tools something to ground in. The same work serves both.

Part of it falls on the profession's training structures, and this is where I expect the most resistance, because it threatens the economics of the pyramid. If juniors are not drafting labour, firms have to decide what they are, and pay for the answer.

The study is usually read as a report on tools. Read it instead as a time-and-motion study of a profession mid-transformation, and the finding is starker than any adoption number: the work that defined the transactional lawyer has already changed hands, the work that remains is judgment, and judgment is currently produced by an apprenticeship system whose raw material is disappearing. The lawyers in this survey have adapted individually. The profession has not begun.

Download the full study here.

‍