Last year we asked the foundational questions: how do you authenticate AI agents, control what they can access, and ensure they act within the right permissions? Teams have started answering those questions. Now, harder ones are surfacing. AI agents are proliferating across your organization's tooling. They hold credentials, access sensitive resources, and act on behalf of users who may have changed roles or left the company entirely. When an employee off-boards, how many orphaned agents keep running with their permissions? When an agent creates a resource, who owns it? When something goes wrong, can your audit trail tell you which agent did what and on whose authority?

This talk goes deeper into the two agent lifecycle patterns IT teams are encountering in practice: short-lived agents that act as scoped proxies for a user, and persistent agents that operate as independent entities within an organization. We will walk through the real governance problems each pattern introduces, including permission sprawl, credential lifecycle, the growing gap between user off-boarding and agent de-provisioning, and the question nobody wants to answer yet: what happens when an agent creates another agent?