Member-only story
The Agentic Runtime Flow Nobody Draws
What actually happens between the question and the answer?
Part 3b of the series “Stop Mixing Things: Understanding the Layers Beneath AI”.
Part 1 covered CDM, Ontology and Knowledge Graphs.
Part 2 covered Ontology, Glossary, MDM, Data Catalog and Semantic Layer.
Part 3a separated find, retrieve, contextualize, connect and act.
You are not a medium member? This is the friend link to read the full blog …
The Diagram Ends Where the Problems Start
Architecture diagrams are great at showing what exists. They’re terrible at showing what happens. You get boxes for the catalog, the vector store, the semantic layer, the graph and the agent, and arrows that suggest they talk to each other. What you don’t get is the sequence: who asks what, in which order, with whose permissions, and what gets left behind afterwards.
For a classic web application, that’s mostly fine. The sequence is boring and well understood. For an AI system it’s the opposite. The interesting decisions, and nearly all the failures, live in the sequence. So let’s draw the thing nobody draws.
