# Retrieval and AI knowledge assistants

> Understand why connecting documents differs from training a model.

Canonical: https://ainextia.com/learn-ai/rag-and-knowledge/
Publisher: https://ainextia.com/

Level: Intermediate
Track: How it works
Reviewed: 2026-11-05

## Part 1
Retrieval-augmented generation first locates relevant documents or passages and then supplies them to a model. It can improve grounding when the source collection is correct, current, and permitted.

## Part 2
A retrieval layer can still select the wrong passage, miss critical context, or leak information if access controls fail. Versioning and citations should be designed at the document level.

## Part 3
Ask a question about a known policy and compare the generated answer to the exact policy version.

## Knowledge check
How should an assistant react when a policy document is absent or contradictory?

Reference: https://learn.microsoft.com/en-us/azure/ai-services/openai/concepts/use-your-data
