Deploying an AI agent and managing one are two completely different jobs. Most of the conversation in CX right now centres on building and launching AI agents. Far less attention goes to what happens after go-live, and that is where most of the risk sits. In this session, Matt will walk through a five-phase lifecycle for managing AI agents in customer support, from onboarding and supervision through to knowing when to retire an agent before it quietly erodes customer trust. He'll cover how to assign ownership, what to monitor and when, how to diagnose performance problems rather than adjust settings speculatively, and how to build the internal discipline that keeps agents operating as intended. Key learnings: - How to onboard an AI agent with defined scope, success criteria, and a handoff protocol from day one - What ongoing supervision looks like in practice, including conversation sampling, metric tracking, and knowledge base maintenance - How to diagnose the cause of performance problems before reaching for a fix - When to retire an agent and how to do it cleanly so you don't lose institutional knowledge