About this session
Models have gotten exponentially smarter — two years ago they couldn't pass the bar, today they score in the top 1%. So why do most agents still fail at a simple business question? The demo works, you deploy it, and the business abandons it a month later. The missing variable is context: the business definitions, procedural knowledge, and operational norms that make a human expert an expert.
Drawing on hundreds of production deployments, Prukalpa Sankar breaks down what it actually takes to give agents contextual intelligence — and get them past the demo stage. She'll walk through the architecture of a context layer: context repos that are versioned, testable, and portable; simulation environments that catch failures before they ship; agent traces that compound back into shared context; and why context engineering scales where fine-tuning and prompting stall. And she'll make the case for why your context has to be open — MCP, Iceberg, deployable to any framework — and what breaks when it isn't.