Log in
← Agentic Way of Work
Investment review

Last deal’s diligence doesn’t help the next one

Run each deal as a scoped project — theses, risks, and open questions recorded as you go — so pattern recognition stops living only in your head.

You reviewed a company like this one two years ago. You remember there was a margin problem, or a channel risk, or a founder question — but the memo is a PDF somewhere and the reasoning behind it is gone. Every deal review starts from a blank page, even when it shouldn't.

Why this keeps happening.

Deal work produces documents — memos, models, call notes — but documents are the output, not the knowledge. The theses you tested, the risks that materialized, the questions that turned out to matter: those live in the reviewer's head, and heads don't scale across a portfolio or a team.

The way of working.

One deal, one project

Each review is a scoped project: thesis, findings, risks, open questions, decision. An Agent keeps the index current, so anyone opening the project sees the state of the review in one screen.

Decisions separated from impressions

A confirmed judgment ("pass — unit economics") is marked as a decision; a hunch stays a labeled assumption. Two years later you know which was which.

The next deal starts with a search

Opening a new review, the Agent searches past deal projects first — comparable companies, sector risks you already mapped, questions that mattered before. Prior work becomes the checklist.

Where this comes from

These are the same mechanics we run daily on our own product work — scoped projects, decision integrity, search-before-work. This article maps them to deal review because that is where several of our early users' workspaces already live.

Start here.

Create one project for a live review and follow Organizing Context. The compounding starts with the second deal.

Bring the method to your work.

Connect from Claude, Codex, ChatGPT, or any MCP client.

Get connected