AI writing assistance, document summarisation, and intelligent formatting suggestions are no longer differentiators – users now expect them. But most implementations share a common architecture: your document content leaves your infrastructure and reaches a commercial API. For public administrations, healthcare bodies, and educational institutions handling sensitive data, that trade-off is often a hard no. The goal is not to resist AI, but to adopt it on terms that do not quietly undo the sovereignty work your organisation has already done.
This session explores how Collabora Office has integrated meaningful AI capabilities, including writing assistance, formula troubleshooting and slide creation, while keeping data firmly within your own boundaries. We examine local and self-hosted LLM integration patterns, the latency and quality trade-offs involved, and the governance questions that AI features introduce even when the model itself is open.
