Member-only story
Stop Mixing RAG, Knowledge Graphs, Semantic Layers and Agents
Retrieval is not understanding, neither is a nice diagram.
Part 3a 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 3b follows with the runtime flow.
You are not a medium member? This is the friend link to read the full blog …
The AI Architecture Whiteboard Disaster
There’s a particular diagram I’ve seen in more AI kick-offs than I can count. In the middle sits a big box labelled “LLM”. Around it, in no particular order, sit “RAG”, “Knowledge Graph”, “Vector DB”, “Semantic Layer”, “Agents” and a cloud that just says “Enterprise Knowledge”. Every box has an arrow to the LLM, and the LLM has an arrow back to every box. It looks impressive. It also tells you absolutely nothing about what happens when a user asks a question.
If you’ve followed this series, you know where I’m going. In Parts 1 and 2 we separated layers by the question each one answers. CDM, ontology, knowledge graph, glossary, MDM, catalog, semantic layer: different questions, different owners, different failure modes. That was the terminology part…
