UTIO
☎️ Try the demoSign up

AI glossary · Data and knowledge

RAG

In one sentence

RAG (retrieval-augmented generation) is a technique in which a language model first retrieves the relevant passages from your own documents and then writes its answer from them, instead of relying on what it memorised in training.

Also known as: retrieval-augmented generation · retrieval augmented generation · knowledge-base AI

How RAG (retrieval-augmented generation) works A five-step flow. 1: a customer asks “Do you ship to Northern Ireland?”. 2: the question is turned into an embedding and searched by meaning across the company's knowledge base. 3: the most relevant passages come back, for example from the shipping policy. 4: the language model writes the answer using only those passages as context. 5: the customer gets the answer together with its source. 1 · Customer question “Do you ship to Northern Ireland?” 2 · Search by meaning Embeddings over your knowledgebase: web pages, PDFs, FAQs 3 · Relevant passages “We ship to Northern Ireland…”Shipping policy, section 4 4 · The LLM writes Using those passages ascontext, not from memory 5 · Answer with its source “Yes, we ship there…”Source: Shipping policy
RAG: the model doesn't answer from memory but from what it finds in your documents, and it can cite where the answer came from.

What is RAG?

A language model knows a lot about a lot, but nothing about your business: not your prices, not your returns policy, not your bank holiday hours. Ask it anyway and it may fill the gap with something that sounds right and isn’t.

Put it to work

AI agents that look after your customers

From €49/mo + VAT

Voice, WhatsApp and email with the phone line included: agent, numbers and CRM in one fee.

See plans and pricing

RAG fixes that with an extra step. Before answering, the system retrieves the passages in your knowledge base that relate to the question and hands them to the model as context. The model then generates its answer from those passages. It answers with your data rather than whatever it remembers from the web, and it can point to the document each answer came from.

Related to this term

UTIO agents

  • Available 24/7 in 26 languages
  • Set it up yourself for €0: built from your website
  • 14-day free trial, no card needed

From €49/mo + VAT

See plans and pricing

How it works

  1. Prepare. Your documents are split into passages and each passage is turned into an embedding: a list of numbers that captures its meaning.
  2. Search. The incoming question is turned into an embedding too, and the passages closest in meaning are retrieved. The words don’t need to match: “do you post to Belfast?” finds the paragraph about “shipping to Northern Ireland”.
  3. Answer. The selected passages are added to the model’s instructions and it writes the reply.

Because only the useful passages travel, each question uses few tokens even if your documentation runs to hundreds of pages.

Example

An online deli in Bristol gets the same questions every day: delivery times, shipping to Northern Ireland, allergens in each product. All of it is already written on its website and in PDF product sheets.

With RAG, the assistant checks those sheets before replying. If a customer asks whether the oat biscuits contain nuts, the answer comes straight from the product sheet. If the documents don’t cover the question, the assistant doesn’t improvise: it says so and passes the query to the team.

RAG at UTIO

UTIO agents answer from your knowledge base: the documentation you load. It’s the core of the support vertical, included from the Business plan: anything that isn’t in the documentation doesn’t get a made-up answer; the agent takes a note and opens a ticket in the CRM with the details already captured.

If you’d rather build your own system, UTIO’s AI API creates embeddings to search your content and turns scanned documents into text with OCR.

RAG is one of the building blocks that turn a general-purpose model into an AI agent that’s genuinely useful to your business.

Frequently asked questions

Does RAG stop AI from making things up?

It cuts hallucinations sharply but doesn't remove them on its own. The agent should also be told not to answer anything it can't find in the documents and to hand over to a person, and the knowledge base has to be kept up to date.

What's the difference between RAG and fine-tuning?

Fine-tuning retrains the model on your data and changes how it writes or reasons. RAG leaves the model alone and hands it the information it needs with each question. For data that changes, like prices or opening hours, RAG is far easier: update the document and you're done.

What documents can a RAG system use?

Almost any text: web pages, PDFs, FAQs, manuals or terms of service. Scanned documents are converted to text with OCR first.

Related terms

  • Token

    A token is the chunk of text a language model works with: a word, part of a word or a punctuation mark. Models read, count and bill in tokens, not in words or characters.

  • AI agent

    An AI agent is a system that uses a language model to understand a request, decide which steps to take and carry them out with real tools, such as your calendar or CRM, until the task is done rather than just replying.

Updated on

← Back to the glossary

UTIO is the AI agent line of Universal Telecom, an operator registered with the CNMC.