Legal case study
An AI document assistant that takes the first pass at contract review
LexMate serves legal teams, law firms and in-house counsel who work through contracts and legal documents at volume.
The problem
First-pass contract review is mechanical work done by expensive people. Someone qualified reads the same clauses again and again to find the handful that depart from the firm's standard position. The queue builds, the business chases, and the real risk is that something gets waved through because there was no time to read it properly.
What we built
- Automated drafting that works from the firm's own templates and precedent
- Contract review that flags where a document departs from the standard position
- Clause extraction, so obligations and terms become searchable rather than buried inside PDFs
- Compliance checking built into the review step
- Natural language processing applied over legal content rather than general-purpose text
What made it hard
Legal AI has an unusually low tolerance for being approximately right. A wrong clause in a contract is not a bad suggestion, it is exposure. The engine had to read legal language precisely while staying adaptable across jurisdictions and document types — and lawyers had to be able to see why it said what it said before they would rely on any of it.
How we built it
- A language model fine-tuned on legal corpora rather than general-purpose text
- A human-in-the-loop review step that flags ambiguous clauses for manual attention instead of deciding quietly
- An explanation attached to every suggestion, so a lawyer can evaluate it and override it
- Version control and audit trails to the standard legal practice management expects
Who it is for
Legal teams, law firms and in-house counsel who work through contracts and legal documents at volume — the people for whom first-pass review is a standing queue rather than an occasional task.
The product
Building something similar?
Thirty minutes with the people who would actually build it. You leave with a scope and a timeline.

