Fintech

A multi-agent diligence engine for a mid-market PE firm

An agent system that ingests a target company's data room and produces a 40-page diligence brief in 9 minutes — versus the team's previous 11-day average.

Brief turnaround
11d → 9m
Analyst trust score
94%
Time saved / yr
$2.1M
Hallucinated cites
0
The challenge

PE diligence is unforgiving — a hallucinated number in an IC memo is a career-ending event. Off-the-shelf RAG demos didn't survive a single real data room.

Our approach

Five specialized agents (financial, legal, customer concentration, market, ops) coordinated by a planner. Every claim cite-bound to source pages. Reviewer queue for the partners.

The outcome

After four months, Mercer used the system on 11 live deals. Analyst time per deal dropped 71% with zero hallucinated citations across 4,400 IC-grade claims.

The cite-back guardrail was the unlock. Once every claim was traceable to a page in the data room, our partners stopped second-guessing the output.

Marcus Whelan · Managing Director, Mercer Capital
Project details
Client
Mercer Capital
Industry
Fintech
Duration
16 weeks
Region
New York, NY
Tech stack
LangGraphPostgres + pgvectorClaude 3.5 SonnetTemporalS3 + KMS
Services delivered
AI Agent DevelopmentRAG SystemsArchitecture Review

Similar challenge?

Let's scope your project

Book a call and tell us what you're trying to ship. A senior engineer will respond within one business day.