Retrieval (RAG)

Retrieval, often called RAG, is a way of grounding an AI in your own material. Instead of relying only on what the model learned in training, the system first searches your documents - a knowledge base, a set of policies, a product catalogue - finds the passages that fit the question, and hands them to the model to answer from. The AI works from your real information rather than its best guess.

Why it matters for business

This is how you build an AI assistant that actually knows your business. A support bot that answers from your help docs, an internal tool that searches your policies, a sales assistant that quotes your real pricing - all use retrieval. It cuts down hallucination sharply because the answer is tied to a source you can check.