What Is RAG? A Knowledge-Base Chatbot Guide
How RAG retrieves information, its benefits and limits, and why it matters for enterprise chatbots.
Short answer
Retrieval-augmented generation finds relevant passages from defined sources before giving them to an AI model as response context.
A simple RAG flow
A document is processed and chunked, the question becomes a search, relevant passages are selected, and the model receives those passages with answer instructions.
Primary benefit
Business knowledge can change independently from model memory, and the retrieved context can be traced to a source.
Limitations
RAG does not guarantee truth. Incomplete sources, weak retrieval, or vague instructions can still produce a poor answer.
- Source quality and freshness
- Chunking
- Result ranking
- Confidence and escalation