Understanding RAG: enriching a model with a controlled corpus.

Retrieval-Augmented Generation (RAG) is an architecture where a system retrieves excerpts from a corpus, then provides them as context to a language model. The goal is not to know everything, but to answer from documents the organisation chose to expose.

When RAG is relevant

It becomes interesting when a useful document fund exists — procedures, files, internal bases — and users lose time searching. It is less relevant without a corpus, clear rights or a business question.

What RAG does not promise

Access rights, citations, evaluation and logs are part of the product. Without them, the system may expose information to the wrong person or create excessive confidence.

Artificial intelligence — To connect this building block to an enterprise AI approach.