Retrieval · Short note

RAG is how the answer stays yours

A model can sound sure about a policy it has never seen. Retrieval is what puts your document in the answer.

A language model answers from what it learned in training, plus whatever you paste into the prompt. Your procedures, your contracts, and last month’s incident notes are not in that training set. Ask it “what is our rule for a missed medication?” and it can still produce a confident paragraph. The rule in that paragraph may belong to nobody you employ.

Retrieval-augmented generation changes the order of the work. You keep the documents. When the question arrives, the system finds the passages that match, and the model writes from those passages. The answer can name the page. A reviewer can open the same page and see whether the sentence is fair.

That is why RAG is the future of answers that have to stay true to a company’s own files. A bigger general model gets better at language. Your files still have to be retrieved, on the day the question is asked, from the copy that is current.

We would rather a short answer with the right paragraph beside it than a long answer from memory. The citation is the part a manager can argue with. Everything else is tone.

This is a short note. The longer essay is still being written.