Analysis
A lawyer opens a new NDA. The system flags terms that depart from the firm’s playbook; the lawyer still has to decide what those differences mean.
An October 28 OpenAI customer story describes how Legora built that playbook from previously signed agreements. The company’s general counsel estimates NDA review fell from around a fifth of the team’s capacity to under one percent. That is a striking shift in time, though it is an estimate within a vendor case study, not an independent measure of accuracy or legal outcomes.
In the same story, Legora’s chief executive cites a Fortune 100 customer reporting 90 percent faster contract turnaround and 9.5 hours returned per lawyer each week. Those figures are customer outcomes presented by a vendor; they do not show how performance was compared or establish controlled error rates.
Legora’s broader claim is about searching large contract collections across multiple questions, with lawyers assessing what the system surfaces. More coverage could help a team find issues it might otherwise miss. But a tidy answer is not the same as a complete search. For a reviewer to trust the handoff, it should show which documents were processed, which questions were answered, and where the system could not reach a conclusion. Each finding should lead back to its source, while an unresolved issue should stay visibly unresolved.
That is the practical test for any agent handling legal work: ask for a sample handoff and trace a finding to the underlying document. If the system cannot show what it checked and what it left open, the review is not finished.
Source: Legora is building the agentic operating system for legal work · 2026-10-28
Why it matters
Business owners evaluating legal AI should ask for evidence of what was reviewed and a clear account of what still needs a lawyer.
