What is RAG (retrieval-augmented generation)?
Retrieval-augmented generation is the pattern of searching a corpus for relevant material and placing it in the model’s context, so its answer is grounded in specific retrieved sources rather than in training data alone.
It addresses staleness and specificity: the model does not need to have memorised your documentation, only to be handed the right part of it at the right moment.
It does not, by itself, address permissions. A retrieval layer that searches everything will happily ground an answer in a document the asker was never allowed to read — which is why retrieval has to be scoped by the same identity as everything else.
Knowledge is scoped to an environment and reviewed, and retrieval happens under the run’s identity rather than as an unscoped search.
Related terms
- AI agent — A system that uses a language model to decide which actions to take, then takes them — as opposed to only producing text.
- Multi-tenancy — One deployment serving many customers, with each customer’s data isolated from the others.
- Audit trail — A durable record of what a system did, who caused it, and what it touched — sufficient to reconstruct events afterwards.
See it working rather than described.
The product demo runs one agent across three systems and stops at the approval gate.