AI agents don’t log in–they execute, chain decisions, and generate access paths no one explicitly approved. In this session, Andrej Safundzic strips away the hype and shows exactly where today’s IAM model collapses under autonomy and what actually replaces it. You’ll learn: Why provisioning and governance are the wrong control points for AI Where risk hides: dynamic access, self-directed workflows, and missing ownership Why visibility without enforcement is just lagging insight How to move to a runtime control model that acts in real time AI isn’t just another identity problem. It requires a new control plane and a shift in how teams approach it.