RAG for companies is an AI assistant that answers questions from your own documents and shows the exact passage behind every answer. I design, build and measure it for companies in Slovakia and Czechia, in the cloud or fully on-premise.
AI that doesn't cite its source is guessing. Every answer I build points to the exact passage in the document.
Who it is for
- Companies that lose time searching through contracts, manuals and policies
- Teams that answer the same customer or colleague questions over and over
- Companies where ChatGPT is not an option because data must not leave the company
If you have a few dozen pages of documents and a question once a week, RAG won't pay off. A well-organized wiki will do.
How it works
- We pick the documents and real questions we will measure quality on.
- I build a prototype on a sample of your documents: hybrid search, a reranker, citations.
- We measure accuracy, and whether the AI declines to answer when the documents don't contain the answer.
- I deploy it, connect access rights and set up ongoing quality measurement.
What you get
Test set
Real questions with correct answers, used to measure every change.
Search with citations
Hybrid search and a reranker. Every answer links to a passage in the document.
Refusal when there is no answer
If the answer is not in the documents, the AI says so instead of making it up.
Access rights
Each person only gets answers from documents they are allowed to see.
Report with numbers
Accuracy before and after, refusal rate and a list of questions where the AI still fails.
In practice
On Construction Graph RAG I raised retrieval accuracy (MRR) from 76.1 to 96.5 % on 128 test questions. 99.2 % of answers contain exactly the expected facts, and the AI correctly declines 90 % of questions that have no answer.
- Construction Graph RAG
99.2 % accurate answers on a local model.
What changes in the company
An answer someone now spends tens of minutes digging out of contracts and manuals arrives in seconds, with a link to the source. Colleagues who keep answering the same questions get back to their own work.
Technology and data security
Documents can stay on your server with a local model, or go through an EU-region API with data retention turned off. Every answer cites its source, so it can be checked.
Price
First version over a smaller document set from €490. Production RAG for a company €3,600–5,400, 3–6 weeks. Pricing
FAQ
What is RAG?
RAG (retrieval-augmented generation) is how AI answers from your documents: it first finds the relevant passages and only then writes an answer from them, citing the source.
Is RAG better than fine-tuning?
For questions over company documents, yes. RAG answers from current documents and cites them. Fine-tuning changes how a model behaves, not what it knows. It makes sense for answer style or format.
Does RAG work with Slovak documents?
Yes. Construction Graph RAG answers Slovak questions from documents in Slovak and English and cites the exact passage.
How much does an AI project cost?
Private AI over your documents starts from €490, an AI audit costs €540–1,080, automating one process €800–2,400, and production RAG or an AI agent prototype €3,600–5,400. My hourly rate is €35–45 and the first 30-minute call is free.
Related
- Services
- Local LLM
AI on your own server. Data stays in the company.
- AI agents
The agent drafts, a person approves.
- RAG workshop
Developer workshop: RAG with evaluation.