
November 23–27, 2026
Managing Partner
Intellectus Advisors
When an AI agent gets something wrong, who notices? Who can intervene? And what happens when the person reviewing its work starts accepting everything it produces? Drawing on research across more than 40 real-world AI workflows and my experience building an AI-native product, I explore the operational gaps that emerge when AI moves into everyday use, human review that becomes a rubber stamp, escalation paths with no clear owner, and oversight costs that never made it into the business case. This session applies those lessons to agentic systems, where AI can take actions as well as generate answers. Through practical examples, I show how teams can decide where agents need approval, when exceptions should reach a person, and how to assess whether oversight is actually working. Attendees will learn how to: • Define an agent’s authority and the decisions that require human approval. • Design review and escalation around risk and team capacity. • Identify signs that human oversight has become ineffective. • Account for the human work required to operate agents at scale.
See every session across the AI Agents Conference November 2026 line-up.
View Agenda