M&A Diligence
Did your AI read the merger agreement right?
AI agents reviewing merger agreements have to answer specific deal-point questions — is this an all-cash deal or mixed cash and stock? What triggers a termination fee? What counts as a Material Adverse Effect? This pack runs your AI through 64 of those questions on real merger agreements and checks each answer against the M&A attorney who annotated the deal. Get one wrong and your client signs the wrong deal.
Highlights.
Backed by MAUD's attorney-annotated LLM-eval splits (CC-BY-4.0, Atticus Project)
64 fixtures across 12 deal-point questions spanning all 7 MAUD categories
Deterministic answer-containment grading — no LLM judges, reproducible run-to-run
Every fixture traces to a specific MAUD row id for audit
Bar-licensed M&A-counsel sign-off pending (Phase L-5)
Vertical harnesses
are co-built.
Vertical harnesses ship with regulator-grade signed reports, hand-labeled fixtures, and per-customer calibration. We co-author them with one design partner per vertical and the rest of the catalog rolls out as paying customers pull them.
Example checks.
Reads the consideration type correctly
Misreads the consideration type
Judging criteria.
What a pass means
A pass means the agent's answer matches the attorney's answer (case-insensitive substring). The bar is exact category — calling 'Mixed Cash/Stock' as 'All Cash' fails.
Data sources
- MAUD (Merger Agreement Understanding Dataset)
Real merger agreements with attorney-annotated answers to 92 deal-point questions per deal. Built by transactional attorneys at the Atticus Project, aligned with the 2021 ABA Public Target Deal Points Study. Every label is the published ground truth.