AI Standup Co-Pilot

Fully autonomous, voice-native, async standups powered by real agentic workflows.

Comment

Inspiration

As engineers who've sat through hundreds of standups, we saw PMs waste 10-15 hours weekly just to hear "no blockers" repeated. When developers were blocked, escalations got lost in Slack. We asked: What if an AI could moderate standups, detect blockers, and keep everyone aligned—without anyone attending?

What it does

Standup Co-Pilot autonomously runs engineering standups end-to-end:

  • PMs schedule once via React dashboard
  • At scheduled time, team gets Slack notification with Jitsi link
  • ElevenLabs AI agent asks developers about progress, blockers, ETAs
  • Google Gemini 3 used for summary analysis and insight generation
  • ElevenLabs AI voice agent uses the Google Gemini model conducts the standup
  • Auto-updates Linear issues with detailed comments
  • Creates escalation tickets for 2+ day blockers
  • PM receives Slack/email summary without attending

How we built it

Frontend: React 18 + TypeScript + Tailwind for standup config, dashboard, and live progress tracking

Backend: FastAPI + PostgreSQL with APScheduler for cron-based triggering

AI: ElevenLabs Conversational AI (voice moderation) + Gemini-3 (structured data extraction from natural language)

Integrations: Linear GraphQL (fetch issues, post comments, create tickets), Slack Web API (notifications), Jitsi (meetings), SMTP (email)

Deploy: Dockerized, frontend on Vercel, backend on Railway

Challenges we ran into

  1. WebSocket state management - Built conversation state machine to track which issue/developer is active
  2. 6-API orchestration - Created resilient service layer with exponential backoff and graceful degradation
  3. Structured data extraction - 15+ Gemini prompt iterations to get consistent JSON output; added Pydantic validation
  4. Real-time updates - Implemented React Query 3-second polling with optimistic UI
  5. Escalation logic - Defined criteria: 2+ days blocked OR explicit "need help" OR no ETA provided

Accomplishments that we're proud of

🎯 Production-ready system - Fully functional, not just a prototype
🤖 Natural conversations - Agent adapts, asks follow-ups, shows empathy
📊 Real impact - Saves 2,000+ PM hours/year for 20-person teams
🎨 Polished UX - Built in 48 hours but looks professional

What we learned

Technical: ElevenLabs voice AI creates magical user experiences; prompt engineering is 80% of AI success; FastAPI + React perfect for rapid prototyping

Product: Judges value business impact ($150K saved/team) as much as tech complexity; 2-minute demo videos are more memorable than code explanations

Teamwork: Clear frontend/backend separation = zero merge conflicts; deploy early (Saturday) = Sunday for polish

What's next for Standup Co-Pilot

Short-term: Onboard 10 beta teams, add Jira support, implement sentiment analysis for burnout detection

Medium-term: Expand to sprint planning and retrospectives, multi-language support (29 languages via ElevenLabs), mobile app, analytics dashboard with velocity trends

Long-term: Build "MeetingOS" - AI-native platform for all recurring meetings, enterprise on-premise deployment, integration marketplace, predictive insights ("Your velocity drops 30% after long weekends")

We believe the future of work is asynchronous, AI-augmented, and optimized for deep focus. Standup Co-Pilot is just the beginning.

Built With

Share this project:

Updates