Episodic logs are not structured state.
Short-term scratchpads, long-term knowledge stores, and session boundaries each solve different problems. Treating chat history as memory is why your agent forgets what mattered.
Short-term scratchpads, long-term knowledge stores, and session boundaries each solve different problems. Treating chat history as memory is why your agent forgets what mattered.
Tabular enterprise data needs schema-aware retrieval, SQL or semantic layers, and explicit join logic. Dumping rows into chunks and embedding them is why agents confidently cite the wrong quarter.
I've replaced three 'RAG is solved' pipelines this year. The pattern is always the same: layout-aware parsing, hybrid retrieval, and a reranker, not a bigger embedding model.