We extract how your firm actually evaluates deals, codify it, and overlay it on the AI you already run.

Everyone runs the same models over the same deal documents. The edge is how your senior people evaluate a deal, and almost none of it is written down: the partners carry it, everyone else approximates it, and the AI stack you just deployed has none of it. We extract that judgment, codify it as a House View File your firm owns, overlay it on your stack, and keep it current. 4 engagements, each standalone, each building on the last.

1. Framework and diagnostic

We decide with you what is worth capturing: where your senior judgment matters most, where an LLM can actually apply it, and what lives only in people’s heads. We measure the baseline in week one by running your own stack on deals you have already decided, so the readout has a control group and no partner time is spent yet. Where useful, the diagnostic also puts a number on the spread between what you say you weight and what your decisions reveal (the free instrument on this site is the 4-minute version). Out of this comes the scope and the one number the engagement is measured on.

2. Extraction

Structured interviews with your senior deal-makers, anchored on specific past deals rather than stated principles: the one that cleared when lookalikes did not, the burn, the pass everyone still argues about. Every adjective gets chased to a number. We show each partner what the model would have said on the same deal and ask what they would change; the reaction extracts what a blank page never does.

A starter House View File comes from one 25-minute interview and ships in 48 hours. The full file covers the partnership.

3. Codification

The House View File: dimensions and weights with calibration history, disqualifying flags with triggers and survivable exceptions, positive signals, anti-patterns, judgment notes on where the rules bend, worked examples, and loading instructions for any stack. Every rule carries the name of the person whose judgment it encodes and the words they used, so it carries weight in a memo and can be audited later.

4. Overlay and calibration

We load the file into the stack you already run, whether that is SharePoint and Copilot, your deal CRM, or internal tools, so documents go in and a structured, attributed read comes out before the Monday meeting. Deliberately not a platform sale: your judgment stays inside your walls, in a file you own.

Then we keep it honest. Quarterly, we review which flags fired, where the lens and the actual decision disagreed and what that revealed (a stale rule, or a right one that got overridden for a reason worth encoding), which dimensions never move, and we revise, re-scoring history so comparability survives the change. Heuristics decay; nobody internal ever schedules the session to refresh them. This is the session.

What never leaves your firm

Interviews operate at the level of heuristics, never holdings. No positions, no terms, no pipeline. Nothing pools across clients: a lens is close to worthless outside the shop that produced it, and the contract says we retain patterns, never data. The trust argument is architectural, not a promise.

Who this is for

Credit funds, buyout firms, growth investors, and independent sponsors screening 100 or more opportunities a year with 5 to 20 investment professionals. If one person evaluates everything personally, you do not have the problem we solve.