Work with your team in Interact
Interact is the cockpit. It is where you ask your team for something, see what it intends to do, watch it work, and step in when you want to.
It is not a chatbot. The distinction matters: a chatbot answers you, and your team does things on your behalf.
Plan first, then work#
Say what you want to be true. “Qualify the new inbound leads from this week” is a better ask than a list of steps, because your team can plan the steps and you cannot always know them.
Anything non-trivial comes back as a plan before it runs: what your team intends to do, in what order, and what it will touch. Nothing has happened yet at this point.
Approving releases the work. If the plan is wrong, say so in the same thread and it re-plans rather than stubbornly retrying.
The activity pane shows the work as it runs, step by step, with each coworker’s contribution as it lands.
The plan step is the cheapest place to catch a misunderstanding. Reading a five-line plan takes ten seconds; unpicking work that ran on the wrong assumption takes considerably longer.
Stopping things#
Two controls, and the difference matters:
- Stop this halts the piece of work you are looking at.
- Stop everything your team is doing halts all in-flight work across the workspace.
Both take effect immediately. Anything already completed stays done, and the run record shows where it was stopped and by whom.
Directing the work#
- Mention a specific coworker with
@to route the ask to them rather than letting the team decide. Useful when you know who owns the area. - Projects group related threads, so a long-running piece of work keeps its context instead of starting cold each time.
- Attachments can be added to an ask, and become part of what your team is reasoning over.
- Check-ins can be scheduled, so a thread comes back to you at a set time rather than you remembering to chase it.
What you will not see#
Interact never shows which underlying model is doing the work, and it does not expose model selection. Your team routes work to the right capability tier itself. If that sounds like a limitation, it is the opposite: it means a model upgrade improves your workspace without you re-configuring anything.
If something goes wrong#
- The plan is repeatedly wrong. The gap is usually context, not capability. Check what your team actually knows in Brain.
- Work stalled mid-run. Check Inbox; a stall is usually a question waiting on you.
- You stopped something by accident. Ask again in the same thread. The context is still there.
- A coworker you mentioned did not respond. Check they are still on your roster in Team.