In credit, risk-taking is often split between two complementary disciplines.
The first is understanding the investment: the business, the capital structure, the documents, and the underwriting.
The second is deciding whether to own it, at what price, and in what size.
Neither discipline is performed by a single person. Credit desks don't rely on a single generalist. Research analysts cover the business. Instrument specialists cover the security. Restructuring experts weigh in when situations become more complex. Portfolio managers synthesize those perspectives into an investment decision.
I built the /eigen orchestrator around the same idea. One command routes work to specialized sub-agents in parallel: instrument specialists, sector specialists, and restructuring specialists when distress signals emerge.
The harder problem isn't building specialized agents. It's orchestrating them around the way a specific investment firm actually makes decisions: its risk thresholds, its heuristics, and its institutional memory across market cycles.
Once calibrated, the system can handle roughly 85% of the repeatable investment process. The remaining 15% belongs to investment professionals: interpreting technicals, assessing relative value, understanding positioning, sizing risk, and deciding whether this is the right investment at this price, at this moment.
eigenCredit is not designed to replace investment professionals. It gives them a better starting point, one that already reflects their firm's investment DNA.
That final 15% isn't a failure of AI. It's the point. LPs allocate capital to investment teams because they trust their judgment and hold them accountable for the outcomes.