Retrieval-augmented generation (RAG) is a technique that grounds a language model’s answers in retrieved documents or data, so responses are based on your own up-to-date information instead of the model’s memory alone.
How it works
When a question comes in, a RAG system first retrieves the most relevant passages from a knowledge source — often using vector search over embeddings — and then passes those passages to the model as context. The model composes its answer from that supplied material, and can cite where each fact came from.
Why it matters for enterprise AI
RAG is how enterprises make general models useful on private, current knowledge without retraining them. It reduces hallucination, keeps answers traceable to a source, and updates instantly as the underlying documents change — the practical foundation for trustworthy AI assistants over company data.
