The challenge
Staff wanted language models for drafting, summarizing, and querying internal documents — but uploading the institution’s files to external cloud services was not acceptable: member records, administrative correspondence, and unpublished documents. The requirement: AI capability, inside the walls.
The solution
We deployed a complete AI environment on the institution’s own hardware:
- A dedicated GPU workstation configured for continuous operation
- Ollama managing local models, selected to match the hardware’s capacity
- Open WebUI as the staff-facing interface with accounts and permissions
- Access control: the service reachable only inside the institution’s network
- Training sessions for staff on effective use and the tool’s limits
The outcome
The institution now has an internal AI assistant staff use daily for drafting and summarizing — without a single document leaving its servers, and without per-user monthly subscriptions.