Company AI knowledge base: your own ChatGPT on your documents
A company AI knowledge base (RAG) is an assistant that answers from your own documents: policies, contracts, manuals, quotes and wikis. Every answer shows its source, access rights are respected, and the data stays in the EU or on your own servers. Staff ask in plain language, in Czech or English.
What companies use a knowledge base for
Anywhere people search documents for answers or ask colleagues the same questions again and again.
Internal helpdesk
Questions on IT, HR, travel expenses or policies. Staff get an answer with a link to the current document.
Onboarding
A new colleague asks about a procedure, a system or a contact without going round the office.
Sales and support
Specifications, price lists, past quotes and answers to common customer questions in one place.
Contracts and documentation
Finding clauses, deadlines and obligations across contracts, with the source for every answer.
Production and service
Manuals, standards and service records. A technician asks from a phone in the field.
A base for AI agents
The same knowledge base then serves agents that answer customers or process documents.
How we deploy it
- Day 1
Consultation
We choose the area, the document sources and who may see what.
- Within 3 working days
Free prototype
An assistant on a sample of your documents and a set of test questions. If you don’t like it, you don’t pay.
- Within 30 days
Live
Sources connected, new documents synced automatically, access rights and training.
- After
Extending
We track what the assistant could not answer, add documents and bring in more departments.
Pay when it works.
Within 3 working days we build a free, working prototype on your own data. If you don’t like it, you don’t pay.
- For fixed-price projects with one clearly defined process
- After the free consultation and a written scope
- We need access to the data and systems named in the scope
- AI model costs during the prototype are on us
Off-the-shelf assistant, your own knowledge base or a fine-tuned model?
Three routes to a “company ChatGPT”. Most companies only need the knowledge base.
| Off-the-shelf assistant (ChatGPT, Copilot) | Own knowledge base (RAG) | Fine-tuned model | |
|---|---|---|---|
| Answers from | General knowledge and files the user shares | Only approved company documents | Learned patterns |
| Source with the answer | Depends on the tool | Always | No |
| New documents | The user uploads or shares them | Synced automatically | Retraining |
| Where it runs | The provider’s cloud | EU cloud or your servers | EU cloud or your servers |
| When to choose it | General work with text | Answers from company knowledge | A specific style or output format |
How we prepare the documents
Answer quality depends on the documents:
- We pick the sources and remove outdated versions
- We split documents into parts that make sense on their own
- We inherit access rights from the source
- We set up automatic syncing of new and changed files
- We build a test set of questions and measure accuracy
Sources we connect
For example:
Related terms
Further reading
Questions
What is RAG?
RAG (retrieval-augmented generation) means the AI first finds the relevant parts of your documents and answers only from them. The model is not trained on your data, so adding a new document is enough for the assistant to know it.
Why isn’t ChatGPT or Copilot enough?
Off-the-shelf tools are good for general work with text. Your own knowledge base makes sense when you need answers only from approved documents, precise access rights, hosting on your own servers, or a connection to company systems and agents.
Will staff see documents they have no access to?
No. The assistant inherits permissions from the source, for example SharePoint, and only answers from documents that person is allowed to read.
How do you stop it from making things up?
The assistant only answers from the passages it finds and shows the source for every answer. When the answer is not in the documents, it says so. We measure accuracy on a set of test questions from your own work.
How much does a company knowledge base cost?
A knowledge base for one area starts at €5,000 excl. VAT and is live within 30 days. Running costs depend on the number of users, documents and the chosen model, and we estimate them upfront. The prototype within 3 working days is free.
Bring us one process or one app idea.
In 30 minutes we tell you whether it is worth handing to an AI agent. Then you get a free working prototype within 3 working days. If you don’t like it, you don’t pay.
- We reply within 24 hours on working days
- The engineer who designs it also builds it
- Write in Czech, English, German, Italian or Turkish
Contact
Write to us
Describe the process or app you want done. We reply within 24 hours on working days, and the prototype within 3 working days is free.