Why OpenAI and Anthropic Won't Win Finance

The guest sees public equities as a strong AI application because relevant data is available, but argues that a durable edge may depend on inputs others lack—such as field research, expert insights, and relationship networks. He also says AI could erode expertise based mainly on being smarter or better read, making unique context and information more valuable differentiators.

The guest predicts that leading investment firms and banks could derive 90% of their enterprise value from software, data, and systems within 10 years, and says firms should capture employees’ expertise in systems they own and operate. A banker’s goal of making his thinking, context, and relationships available to junior colleagues illustrates how those systems could scale individual expertise.

The guest sees private markets as an opportunity for AI because coordination, due diligence, and transactions remain largely manual, while unstructured data has impeded standardization. He argues that standardization can support more transaction volume and liquidity, and describes agent-led infrastructure that could assess portfolio-company sales in minutes rather than months, though the pace of change is uncertain.

The guest argues that AI in finance must be auditable, not merely accurate: firms need to preserve how outputs were produced, inspect assumptions and sources, and manage material nonpublic information appropriately. Those controls become more important as AI moves from copilots to agents that make decisions or take actions and must be explainable to users and regulators.