Today we are opening Workspace with the piece everything else builds on: your assistant. It is not a chatbot bolted onto a productivity tool. It reads your work, remembers your decisions, and can take action across the modules you have connected—inside the permissions you already set.
Grounded, not generic
A generic model starts from zero every time. The assistant combines three sources so its answers fit your org: your conversation history, your knowledge base, and your live workspace context—the project you are in, the schedule ahead, the record on screen. You explain once; it carries the thread forward.
Plan before it acts
Speed without control is a liability in regulated work. Flip the Agent / Plan toggle and the assistant proposes a read-only plan first—nothing changes until you approve it. When it does act, every step is scoped to what the invoking user is allowed to do.
Pick the model that fits
The model selector offers eleven providers across US, EU, and Asia regions, so you can match reasoning, latency, and data-residency needs per conversation. Where your data is processed is a procurement question, not an afterthought—see the Trust Center for residency and subprocessors.
This is the foundation the rest of Workspace plugs into. Book a demo to see the assistant grounded in your modules.