Most organisations are discovering that agentic AI is harder to get right than the pitch suggested. Agents make decisions that seem reasonable to a model but cause real problems in practice - not because the technology is flawed, but because it is operating without sufficient context. Without a clear picture of your software estate - who owns what, what depends on what, and what the rules are - autonomous agents are unpredictable.

This session is for the platform engineers who have spent years building the foundations that matter when it comes to making agentic AI work. The context layer, the ownership model, the dependency graph - these are not new concepts. They are the pillars platform engineers have been building for years.

We will explore why context is the critical missing ingredient in most agentic deployments, and demonstrate what it looks like when agents have the organisational awareness to act correctly - and when humans retain the visibility to know that they have.