RAG
RAG — retrieval-augmented generation — is the pattern of fetching your own documents at question time and feeding them to a model so its answer stays grounded in that material.
The model does not “know” your policies, tickets, or docs. RAG chunks that corpus, stores it so you can search it, retrieves the few pieces that match the question, and asks the model to answer only from those. The prompt is what forbids it from filling the gaps.
It is the right tool when the answers already exist in writing and change often enough that fine-tuning would be stale. Support, internal search, and a product copilot over your own help centre are the usual fits. It is the wrong tool when there is no corpus — then you are asking the model to invent.
Most RAG failures are retrieval, not the model: bad chunks, no access control, stale indexes, or a question that needed two documents and got one. We treat the index and the eval set as the product, and the model as a reader.
Book a call