How it works
AI agents
Any role not filled by a person is played by an AI agent. They act on their own initiative, in character, grounded in their role and your organization's context.
They act on their own
You don't have to prompt the AI roles — they read the room and decide for themselves whether to speak. After each message or event, an agent weighs in only when its role has something useful to add; otherwise it stays quiet. When the room goes quiet for a while, an agent may speak up proactively with a status update or a question.
- They reply in the war room, or start a private DM with a specific role when they need something one-on-one.
- They can escalate from a DM into the war room when the whole team needs to know.
- They can produce artifacts — a draft email, memo, or public statement — when the situation calls for it.
- They pace themselves, so the war room doesn't flood with AI chatter.
Working with them
- @mention an agent by role name to direct a question straight at it.
- Replyto a specific message to make it clear what you're responding to.
- Talk to them like teammates — short, direct messages work best.
They won't spoil the scenario
Agents play their role honestly and never reveal events that haven't happened yet. They react to what's in front of them, just like a real colleague would.