Explainer
RAG: how to teach AI your documents.
RAG (Retrieval-Augmented Generation) is a technique that lets AI answer from your own documents: contracts, policies, price lists or manuals. Before the model writes an answer, it first looks up what is relevant in your material.
In short
- AI searches your documents first, then answers – and shows the source.
- There is no need to “retrain” the model; it just needs access to the documents.
- The quality of the answers depends on how good and up to date the documents are.
On this page
The principle: search first, then write
Imagine a new colleague who, for every question, first opens the company folder, finds the right document and only then answers. That is exactly how RAG works:
- Retrieval – the system finds passages in your documents that relate to the question.
- Augmentation – it passes these passages to the language model together with the question.
- Answer – the model answers from the text it found and states where it came from.
Why RAG matters for businesses
- Less making things up. The model answers from your data, not “off the top of its head”, which greatly reduces the risk of hallucinations.
- Up to date. Change the price list and from that moment the agent answers according to the new one. Nothing needs retraining.
- Traceability. Every answer shows its source document, so checking is quick.
- Access control. The agent sees only the documents you give it access to.
Examples of use
- Law firm: searching case files and contract templates.
- Municipality: answering citizens from approved information on local fees and forms.
- Online shop: answering questions about products and terms and conditions.
- Manufacturer: finding procedures in technical documentation.
You will find more scenarios in the agent catalogue.
What to watch out for
RAG is only as good as your documents. If you have three versions of your price list and two contradict each other, AI will not sort that out for you. Before going live, it pays to review and consolidate the documents and decide who will keep them up to date.
The second thing is personal and sensitive data. Include only what the agent genuinely needs. More in AI and GDPR.
Frequently asked questions
Does the AI model have to be trained on my documents?
No. With RAG the model is not retrained. The documents stay in a search database and the model simply looks into them when it answers.
Which documents are suitable?
Any text that can be read: PDFs, Word files, web pages, emails, spreadsheets, internal wikis. Scanned documents first need to be converted to text.
Will the agent see all our documents?
Only those you give it access to. Access can be limited by folder, by document type and by who is asking the agent.
Want to know where AI would save you time?
We go through a normal week with you and find the first task worth handing to an agent. No obligation.
Related articles
AI hallucinations: why AI makes things up
Why AI sometimes states something untrue with full confidence, and how to handle it in a business: your own documents, cited sources and human checks.
→Explainer · 3 minLarge language model (LLM)
What a large language model (LLM) is, how it works and why it powers both ChatGPT and AI agents. A maths-free explanation for business owners.
→Overview · 3 minAI and GDPR: what a business must watch out for
A practical guide for businesses: legal basis, processing agreements, data transfers outside the EU, automated decisions and DPIAs when deploying AI.
→