What is RAG in AI?
RAG (retrieval-augmented generation) is a technique in which an AI system first searches your own documents for relevant passages and then has a language model write the answer from them.
Also calledretrieval-augmented generation
Why it matters for a business
RAG is usually a quicker route than retraining a model to an AI assistant that knows your products, policies and contracts.
More
The term comes from a 2020 research paper by Lewis and colleagues at Facebook AI Research.
Answers can cite the source document, so staff can check them. When a document changes, you update the index, not the model.
How it works
Documents are split into chunks, stored with embeddings, and the closest chunks are added to the prompt for each question.
RAG vs fine-tuning
RAG gives the model facts at question time; fine-tuning changes how the model writes, not what it knows today.
Quality starts upstream
Answers are only as good as the retrieval: outdated, duplicate or badly scanned documents produce wrong answers.
- InputA question, a message, a document
- RAGContextYour data, found by search
- Language modelReads, reasons and writes
- ResultAn answer, or an action in your tools
What it costs
Starting prices of our services on this topic, in your currency.
RAG in practice
A way to make an AI model answer from your own documents by fetching the relevant passages first.
Example
- A software reseller in Reading has years of support tickets and product manuals.
- With RAG, its support assistant finds the matching manual section and past fixes for each new ticket and drafts a reply with links to them.
- Support staff review the draft before it goes out.
Which term is it?
A large language model: an AI model trained on huge amounts of text to read and write language.

From the founder
I want AI that does a real job inside your tools, with a person checking what matters. We test it on your own data before you rely on it.
Anatoly MikhailovFounder of Pazl







