This executive workshop is for compliance, risk, legal, and technology leaders navigating the next phase of agentic AI adoption, such as moving from experimentation to governed deployment in regulated financial workflows.
It focuses on a practical question for Compliance Officers: how does a firm prove what an AI agent was allowed to see, what it actually used, what policies constrained it, when a human intervened, and whether the decision can be reconstructed later?
Attendees will explore how core standards (such as fiduciary duty, Reg BI, MNPI restrictions, and human oversight) apply when agents retrieve data, reason across sources, and generate recommendations and surveillance rationale.
The live demonstration
A live demonstration walks through a governed surveillance and recommendation workflow:
- An AI agent operating inside an approved data universe;
- Blocked from prohibited sources;
- Constrained by behavioral specifications;
- Routed through human escalation when required; and
- Captured in an exam-ready decision record.
Agenda
Arrival and Registration
Welcome and Opening Remarks
Keynote & Live workthrough: “The Chief Compliance Officer’s Agentic AI Problem”
Interactive Panel: “What would make a Chief Compliance Officer comfortable saying yes to Agentic AI?”
Closing Remarks & Next Steps
Networking Reception
This workshop is designed for:
Why executives attend
- A curated room of COOs, CTOs, and Heads of Data from Fortune 500 wealth management, private banking and capital markets.
- Co-hosted with AWS, the platform these regulated workflows run on, with AWS Financial Services leadership in the room.
- It’s built around the standards you’re actually held to: Reg BI, MNPI, fiduciary duty, and the prudent-investor rule.
You’ll leave knowing how to
- Turn “human in the loop” from a checkbox into a real, auditable control and prove it held.
- Stop the mosaic theory from becoming a data-permissioning failure once agents combine facts at machine speed.
- Translate written compliance policy into machine-enforceable agent behavior with a working technique, not slideware.
- Stand up an agentic control plane that enforces approved data use, blocks prohibited sources, runs policy checks, routes human escalation, and leaves an exam-ready decision record.