#12 — Semantic Layers, Context Layers, and Agents That Stop Guessing

If you work in data, "semantic layer" means one precise thing: the place where "active customer" or "recurring revenue" gets defined once so every dashboard agrees. If you build agents, the word starts doing two jobs at once — and a second term, "context layer," shows up claiming the same ground. This episode of Agentic AI in DevOps picks up directly where episode #11 left off, with Andrey Devyatkin, Vladimir Samoylov, and Fernando Gonçalves untangling the two. The short version: a semantic layer tells an agent what your data means; a context layer tells it what's actually running, where it lives, and what's connected to what. The hosts argue you usually want both — and along the way they get blunt about vendors who promise you can fire your analytics team, about agent memory that "comes stale" and quietly biases every session, and about a Kent Beck line that sums up the whole moment: typing got faster, thinking did not.
Episode page (show notes and links): https://getboris.ai/insights/012-semantic-layers-context-layers-and-agents-that-stop-guessing/