Most AI conversations in investing stop at documents. Upload the CIM. Upload the credit agreement. Upload the transcript. Ask the model what it says.
What's harder to encode is everything the firm has learned making those decisions.
"We don't buy CCCs."
"We've been burned by this sponsor before."
"We need a much higher return for businesses with this capital structure."
None of those live in any document. They emerge over years of investment committee debates, post-mortems, and hundreds of decisions: heuristics, scar tissue, sponsor relationships, sizing calls, risk tolerance, covenant interpretation. That's the accumulated knowledge of the firm, and it's also fragile. It walks out the door when the senior person retires, changes seats, or simply isn't in the room that day.
Two firms can read the same credit agreement and reach opposite conclusions. Not because one reads the document better. Because they decide differently, and until now that difference has lived only in people's heads.
The opportunity in AI for investing isn't a faster reader of documents. It's infrastructure for institutional knowledge, a system built to capture what a firm has learned over decades so it compounds instead of retiring with whoever had it.
The goal isn't an AI that reads documents faster. It's one that understands how your firm thinks.