Onboarding & Mitarbeiter-Suche
New staff ask "How do I book vacation?", "Who is the contact for DATEV?" — and get immediate answers from the employee handbook. Reduces onboarding time by 30–50 %.
Staff spend hours daily searching. A RAG system answers in seconds — from SharePoint, Confluence, Drive, or your document base. With source citation. GDPR-compliant. EU-hosted.
New staff ask "How do I book vacation?", "Who is the contact for DATEV?" — and get immediate answers from the employee handbook. Reduces onboarding time by 30–50 %.
Sales reps ask before a customer meeting "What did we last discuss with Müller GmbH?" — and get a summary from CRM notes, emails, and contract documents.
"Which T&Cs version applied to order X?", "What is in the data protection annex of contract Y?" — answers from the legal contract repository, with source citation and clause reference. Audit log included.
Which systems, which permissions, which data volume.
Embedding model, vector DB, retrieval strategy, permission layer.
Connectors, initial indexing, UI, eval set for quality tests.
Team training, re-indexing schedule, 90-day bug-fix guarantee.
Your documents (PDFs, Confluence pages, SharePoint files) are converted to semantic vector embeddings and stored in a vector DB. For a query, the system searches the most relevant document chunks and answers the question based on them — with precise source citation.
Yes. Embeddings can be generated locally (open-source models), vector DB runs EU-hosted (Hetzner Frankfurt) or on-premise. Model providers only see the relevant question + retrieved snippets, never your complete document base. DPA per component.
SharePoint, Confluence, Google Drive, Notion, local file shares, databases. There are connectors for every source — new sources are integrated in 1–2 days.
Typical ranges: €24,000–80,000 initial build (depending on data volume and number of sources), €60–400/month hosting (embeddings + vector DB + LLM API). Fixed price from Phase II after discovery.
With clean data and well-configured retrieval, typically 85–95 % correct. We always deliver source citations — the employee can check the answer against the original. In disputed cases, the source document is just one click away.
15 min, honest, no sales theater. We clarify whether your document base is RAG-suitable — and in what order of magnitude the build pays off.