Kensink Labs
INTERNAL COPILOTLEGAL TECH · INTERNAL OPSPLACEHOLDER CASE · RIGHTS PENDING

Affidavit Mapp
Internal-research copilot.
4 hrs/analyst saved.

A retrieval copilot threaded into Affidavit Mapp's research workflow. Saved 4 hours per analyst per week within 30 days of rollout — measured at the team level, not estimated.

4 hrs
Saved per analyst per week (n=42)
92%
Retrieval precision on top-3 results
8.4×
Faster than the previous Notion search
30 d
From rollout to measured ROI
01 · THE PROBLEM

Where they were stuck.

Affidavit Mapp's 50-person research team was re-researching the same case law every quarter. Senior analysts complained that institutional knowledge lived in Slack threads and individual Notion pages, none of which surfaced when a junior analyst was assigned a new matter. Time was being lost not to research itself but to finding what had already been researched.

02 · OUR APPROACH

How we built it.

  • 01pgvector retrieval over the firm's case archive + every Notion page tagged 'research'
  • 02Hybrid sparse/dense search with citation-aware reranking
  • 03Slack-bot interface — answers in the channel where work actually happens, with inline source links
  • 04Per-matter context — the bot knows which matter the analyst is currently assigned
  • 05Eval pipeline: 'does the cited source actually support the answer?' measured weekly
I stopped writing matter briefs from a blank page. The bot finds the three most relevant precedents in our archive in seconds — and links me to the original Notion page so I can verify the citation myself.
Research Director
Affidavit Mapp
[TECH STACK]
  • pgvector
  • Next.js
  • Slack API
  • Notion API
  • OpenAI
[ENGAGEMENT]
Duration10 weeks
ShapeINTERNAL COPILOT
HandoffFull ownership
Warranty90 days