Zoë Van Noppen

Building agents is the part everyone talks about. Testing and monitoring is the part that is rarely covered in demos and tutorials. In most teams, checking agent quality looks like this: run the agent, read the output, fix what looks wrong, repeat. Each fix can reintroduce an old bug or a new issue. This session covers the full lifecycle of getting an agent to production and keeping it there: how you systematically test a new version before it reaches users, how you monitor quality once it does, how you manage reliability and governance when the underlying models change, and how you handle versioning and the promotion lifecycle. We will even take this one step further by discussing how you can build a self-evolving test harness. We'll use practical examples, so you can apply the same patterns when building your own agents, or even for your coding agents.