Glossary · Artificial intelligence

What is RAG

RAG (retrieval-augmented generation) is the technique in which the AI searches for relevant passages in a company knowledge base — policies, prices, FAQs — and answers based on them. It reduces made-up answers and lets you update what the AI knows without training a new model.

It's what's behind a support agent that answers about your delivery time, not just any delivery time.

How it works

The customer's question comes in, the system searches the knowledge base for the passages most similar to it and hands those passages to the language model along with the question. The model writes the answer using what it received.

Quality depends on the knowledge base: outdated or contradictory information becomes an outdated or contradictory answer.

Where it's used in sales and customer service

In any answer that depends on the company's own information:

  • Questions about product, terms and timelines in customer service
  • The sales agent's answers during qualification
  • Support for sales reps, who can look up policies and materials without leaving the conversation

Frequently asked questions

Does RAG eliminate made-up answers?

It reduces them, but doesn't eliminate them. That's why the agent also needs a rule to say it doesn't know and hand off to a person.

Do I need to train my own model to use RAG?

No. RAG uses an existing model and only changes what it looks up before answering.

Want an agent that answers with your company's information?

Nexo's free assessment shows what's worth automating first.