Agents now make decisions on your behalf, and you are still the one held accountable for them. Google launched Antigravity by telling developers they had become managers of agents. Kevlin Henney spotted the catch: managers have always been one step back from the work, and that distance is exactly how understanding goes shallow. In a traditional team someone still held the theory of the system. When everyone steps back, nobody may.

That loss has names. Cognitive debt is the gap between what the system does and what the team understands. Intent debt is the why that was never written down. When agents do so much of the work, both pile up faster than ever, and quietly, because every check still goes green. Nobody owns that debt by default. This day is about taking up that lead.

This is AI literacy in its most practical form: reading decisions, not just code, and putting mechanisms into your agentic systems so people can stay answerable. You work on your own pull requests and systems throughout. The spine is Decision4, from my forthcoming book Decision4: A Handbook of Decision Sovereignty for Agentic Engineering. It is one question asked of every decision: if this turns out to be wrong, what does it take to put right?

Any agent tool, with demos in Claude Code using the open-source ai-literacy-superpowers sentinels. It is not a compliance course: it gives organisations the practical end of the AI literacy they are now expected to build. Best for groups of 12 to 24. It also runs as day two of the two-day bootcamp, after How to Build a Lights-On Agentic Factory.