An architecture where a generative model is conditioned on external retrieved documents, typically via vector search, to produce grounded, context-aware outputs.
Why this matters:
If the retrieval layer omits or misranks corporate data, LLMs will answer using third-party sources, creating revenue leakage and brand misrepresentation.