I love AI agents. Really. They write my features, fix my bugs, and open pull requests while I'm still making coffee! It’s amazing! It’s the future! But it's also how I ended up spending two hours on a bug in code that I had technically "written" three weeks earlier.

The thing nobody puts on the keynote slide is that writing code was never the hard part. Owning it is. When production falls over, you have to fix it. "The AI did it" is not an acceptable incident report. (I checked.)

So, let's stop arguing about whether to use agents and start talking about where to use them. In this session, I'll introduce you to the AI Delegation Matrix, a simple framework that categorizes every part of your system into one of four zones. Some code the agent can own outright. Some code the agent can build with your approval. Some code you build together. Other code, such as your payment engine, the agent can examine, explain, and test, but humans will write it. Each zone has real guardrails that exist in your repository, not on a wiki page that no one reads.

Because slides alone have never convinced anyone, we'll put it all to the test in a live demo. What could possibly go wrong?

You’ll leave with a matrix you can apply to your own codebase and a guardrail blueprint for every zone.