# Capstone facilitator notes

**Workshop length:** 2.5 hours (with break).  
**Rooms:** One room with projector; breakout pairs for 25 minutes.

## Agenda

| Time | Activity |
|------|----------|
| 0:00 | Scenario pitch; clarify advisory-only AI boundary |
| 0:15 | Individual read of mock requirements + CSV |
| 0:30 | Breakout: risk register rows H-001–H-003 completion |
| 0:55 | Plenary: regulatory pathway assumptions (Canada first) |
| 1:15 | Breakout: architecture sketch (on-device vs cloud; Canada data residency) |
| 1:40 | Plenary: economics sanity check ($12/patient) |
| 2:00 | Rubric self-score + reveal of sample weak/strong patterns |
| 2:20 | Capture actions; assign owners |

## Common failure modes

- Teams ignore **stale data** UX and only talk about ML accuracy.
- Teams propose U.S. training data flows despite constraint #3.
- Teams omit **PCCP / change control** when discussing quarterly retraining.
- Teams forget **custodian PIA** and affiliate language for Alberta pilot.

## Materials checklist

- Printed rubric or projected from capstone page
- Risk matrix CSV template
- Whiteboard for data-flow diagram