In short
Letting a model look things up before it answers: your documents, your data, the current facts. Grounded answers instead of confident guesses.
A model on its own knows what it learned in training, which is frozen and general. Retrieval-augmented generation puts the relevant material in front of it first: the client's documents, the current numbers, the specific policy. The model then answers from what it was shown.
In practice this is how an agent stays accurate about your business. Before it drafts, it reads. Before it classifies a lead, it retrieves the profile. The instruction file says what to look up and where.
The risk it addresses is invention. A model asked a question it cannot see the answer to will produce a plausible one. Retrieval replaces plausible with sourced.
Example
What it looks like.
Before writing a prospect brief, the agent pulls the company's record, the last three emails, and the relevant product notes, and writes only from those.Related
Read next.
- context windowEverything the model can see on one run: the instructions, the files it was given, the conversation so far. Outside the window, it does not exist.Read the definition
- modelThe intelligence an agent calls when it needs to read, write, or decide. Replaceable. When a better one ships, the same instructions run on it and the results improve.Read the definition
- documentationThe written record of how a system works. Here, the agent keeps it current, so the person who was not in the room can take over.Read the definition