​From RAG to Knowledge Graphs in 2027

Production RAG runs into the same walls: re-chunking and re-embedding on every knowledge update, embedding models that miss semantic nuance, and recursive retrieval that gets expensive fast (have you seen Claude Code?). Knowledge graphs are becoming the semantic layer for agentic workloads (a queryable model of entities and relationships that LLMs consume directly). This talk covers also economics of the migration (one-time knowledge extraction against per-query retrieval over time), where the graph sits in a company’s stack, and what teams get wrong on the way there.