Local AI for office and administration: your documents stay in-house.

In almost every business, knowledge sits in documents: quotes from past years, supplier contracts, machine manuals, minutes, internal rules, email histories. People usually search them by hand, and often only one person knows where things are.

A local AI makes this knowledge searchable like a conversation. Instead of entering keywords, you ask in normal sentences: “What payment terms did we agree with supplier Y?” or “Summarise the changes in the new framework agreement compared with the old one.” The AI finds the relevant passages, answers and names the document the answer comes from.

How it works technically

• Reading in documents: PDFs, Word, Excel, emails and scanned papers (via text recognition) are split into small sections.
• Search index: For each section, a kind of content fingerprint is calculated and stored in a database on your server. This way the search also finds passages that use different words for the same thing.
• Answer: The language model receives the passages found and writes the answer from them, with source references. The technical term for this is RAG (Retrieval-Augmented Generation).
• Permissions: Everyone sees only the documents they would otherwise have access to, for example by department or project.

Who it suits and who it does not

Suitable for offices, trades, retailers, property management firms, agencies, engineering firms, associations and administrations that work a lot with their own documents and do not want to upload them to a cloud provider.

Not offered: The system is not offered for processing health or patient data. This area has special requirements that we do not cover.

What it costs and what you need

For an office of 5 to 20 people, a single computer with a powerful graphics card or plenty of shared memory is usually enough. The software is open source and free of licence costs. So there is a one-off cost for hardware and setup, and afterwards no fee per request or user. Which models run on which hardware is explained on the Technologies page.