Services
From the first question to production.
We run language models on your own hardware, in your building, and use them to automate recurring office work. Consulting, pilot, implementation and operation come from one source, at fixed prices instead of open-ended hours.
Approach
Four stages, each with its own result.
-
Workshop.
We test 2–3 workflows directly on your files, live and locally. You see what a local model can do with your documents and where its limits are.
- Result
- A prioritised list of use cases and a fixed-price offer for the pilot
- Duration
- half a day, on site
- Price
- flat fee, net
-
Pilot.
One automation and the AI workspace for 5–10 testers, on hardware in your building. We measure how long the workflow takes before and after.
- Result
- Go or no-go based on measured figures
- Duration
- 4–6 weeks
- Price
- fixed price, net
-
Rollout.
Set up the server, SSO sign-in, and bring the knowledge base and the automations into production. Infrastructure as code and an operations manual are included.
- Result
- The system in regular operation, documented and accepted
- Duration
- depends on scope, fixed in the offer
- Price
- fixed project price, net
-
Operation.
Updates, model changes and monitoring. If you prefer, we hand operation over to your IT.
- Result
- A system that stays current and monitored
- Duration
- ongoing
- Price
- monthly fee, net
Every stage has a clear result and a fixed price. You decide again after each stage.
Architecture
One server, clear layers.
- Workspace
- Open WebUI in the browser: chat, documents and knowledge bases by group.
- Automation
- Recurring workflows, each approved by a person.
- AI gateway
- One interface for all applications; switching models needs no rebuild.
- Knowledge
- Documents on the file server become searchable (RAG); the originals stay where they are.
- Inference
- Language models on the graphics card; we choose the software for it by measurement.
- Operation
- Monitoring, alerts and backup, coordinated with your IT service provider.
Only established open-source software. Operation can be handed over to another provider at any time.
Security
Security and data protection are built in.
Technical
- Users reach it only via HTTPS on the internal network, not from the internet
- Sign-in via Entra ID or Active Directory, permissions by group
- No open internet access for the server
- File server connected read-only
Organisational
- GDPR: data processing agreement (DPA) and technical and organisational measures (TOMs) as templates
- AI Act: human oversight and documentation are built in
- Restore tested at acceptance
- A non-disclosure agreement (NDA) before the first meeting if you wish
Training
AI literacy for your team.
Since February 2025, companies that use AI must ensure their staff have sufficient AI literacy (AI Act, Art. 4). We train your staff on the system they work with: what the AI can do, where its limits are and how to check its results.
- Format
- on site, on your own system
- Price
- flat fee per session, net
Hardware
You buy the hardware yourself.
Guide prices for the complete system, one-off, net. Which tier fits is something we work out in the workshop.
Entry
from ≈ €4,000
one-off, net. A workstation for small teams and first trials.
Medium
€10,000–12,000
one-off, net. 1× RTX PRO 5000 (48 GB) for 20–30 workstations.
High
from ≈ €18,000
one-off, net. One or two RTX PRO 6000 (96 GB each) for many simultaneous users. Estimated.
Hardware resale
We do not sell hardware. You buy through your IT service provider; we advise on sizing and set the system up.
Next step
It starts with the workshop.
Half a day on your real workflows. Afterwards you know which workflow is worth it and what the pilot costs.