AI Meetings
AI supervisors auto-convene meetings, employees speak in turn, decisions + action items auto-extracted.
Supervisor receives command → auto-decides to convene → creates Meeting + immediate live → each employee LLM speaks around agenda → supervisor wraps up → customer one-clicks to extract decisions + action items → assigned as Missions.
When supervisors auto-convene
- Cross-functional missions (e.g., SEO + content + CS all participate)
- Major decisions (e.g., "brand repositioning")
- Mission Review milestones (DoD complete review)
- Customer explicitly says "convene a meeting"
- Crisis events (e.g., "traffic dropped 20%")
Meeting lifecycle
scheduled → live → ended
- Supervisor opening (LLM generates 1-2 sentences setting topic)
- Each AI employee speaks in turn (around agenda, 1-3 sentences, with opinion)
- Supervisor wrap-up (summarize decisions)
- Customer clicks "AI Extract" → auto-extracted decisions + action items
- Customer one-click "Assign" → action items become Missions
MeetingRoom immersive view
Visit /meetings/<id>:
- Cream warm background
- Participant avatars cluster into a ring, status rings pulse
- Right pane: real-time transcript stream (5s polling)
- Bottom: decisions + action items + one-click assign button
- Customer can manually add a transcript line (POST /transcript)
Permission model
| visibility | Who can see | Typical scenario |
|---|---|---|
| public | Entire workspace | Standup / retro |
| invited_only | convener + invited list | Supervisor decision meeting |
| private | Only participants | 1:1 / pay discussion (forced private) |