In May, Anthropic shipped a KYC screening agent inside Claude for Financial Services. It reads investor documents, checks them against policy, and assembles a packet for a compliance officer to review. It's real, useful work, and a great step forward for the compliance industry.

But just like an engine is not a car, this is only the beginning. The cognitive capability Anthropic delivered is just one component of 11 that an agentic compliance program actually requires. Without the whole system, compliance officers can’t trust the results, AI deployments stay proof-of-concept forever, never delivering the business value they are capable of.

We wrote this to help GPs see the whole picture. The cognition today's AI tools provide is genuinely useful — but it is one of eleven components a compliance program actually needs. If your firm is drawn to AI for compliance, this paper is a guide to the other ten: what to look for, what to ask a vendor, and what has to be true before you trust an agent with your investors' data.