AI contract review that cut M&A due diligence prep from 5 days to 8 hours

Industry

Legal

Market

USA

Client

US law firm with an M&A practice

Sponsor

Partner

What we built

AI contract review

The situation

In an M&A deal, due diligence means reading the target's contracts: customer agreements, supplier agreements, leases, employment contracts, IP licences. Most of each document is standard. The part that matters is a handful of clauses, such as change of control, assignment, exclusivity, termination, indemnity and liability caps, and whether they deviate from what the client would accept.

For the partner at this US firm, contract review had become the biggest cost center: senior lawyers reading boilerplate at $400 an hour to find the clauses that mattered. Preparing a due diligence review took five days, and much of that time was reading rather than judgement.

What we built

An AI review pipeline that does the reading and hands lawyers a brief to review, not a stack to read.

Ingestion. Contracts arrive in whatever form the data room provides: native files, scans, amendments. The pipeline extracts the text, identifies the document type and parties, and links amendments to the agreements they modify.

Clause extraction. For each contract, the AI locates the clauses the deal team cares about and extracts them verbatim, with their location in the source document. Nothing is summarised without the original text alongside it.

Deviation flagging. Each extracted clause is compared with the firm's positions for that deal type: what is standard, what is acceptable, what needs attention. Deviations are flagged with the reason, so a change-of-control clause requiring counterparty consent stands out immediately.

The review brief. The output is the document the deal team would have produced by hand: contract by contract, the key clauses, the flags, and the questions to raise, every point linked back to the page and paragraph it came from. Lawyers verify rather than search.

Lawyers stay in charge. The brief is prepared for review. A lawyer confirms each flag, adds judgement and signs off. The AI never gives the client advice; it gives the lawyers their time back.

Confidentiality by design. Deal documents are some of the most sensitive material a firm holds. Processing runs in an environment the firm controls, with access limited to the deal team and no client documents used to train shared models.

How it went live

We started from the firm's own review checklists and past briefs, so the AI's output matched the format the partners already trusted. The first matters ran side by side with the traditional process, with associates checking every extraction and flag against the source until the firm was confident in the accuracy.

Results

Due diligence prep that took five days now takes eight hours. Senior lawyers spend their time on the flagged clauses and the judgement calls, not on reading boilerplate at partner rates.

What made the difference

Citations on everything. Lawyers will only rely on output they can verify in seconds. Linking every point to its source is what made the brief usable.

The firm's positions, not generic ones. Flagging against the firm's own playbook is what turned extraction into review.

Built around the lawyer's sign-off. Designing the AI as preparation for legal judgement, rather than a substitute for it, is what made the partners adopt it.