Case study · FDE · 30 min
Change Management & User Training for AI
Drive adoption of a new AI assistant with role-based training, champions, and feedback loops — not just a launch email.
FDEtrainingadoption
Interview opener
Ship adoption, not just software.
- 1.Power users vs average users — different plans?
- 2.What does day-30 support look like?
- 3.How do you detect shadow abandonment?
Strong signals
Hit 0/3
Weak signals (avoid)
- One webinar and done
- No adoption owner
Requirements
- Roll out to 500 employees across support and ops
- Target: 40% WAU in 4 weeks; support ticket deflection +15%
- Constraints: 45-min max training slots; mixed skill levels
- Deliverables: curriculum, office hours, cheat sheets, office hours rota
Back-of-envelope Estimation
| Metric | Calculation | Result |
|---|---|---|
| Users | 500 | 500 |
| Training cohorts | 500 / 25 | 20 sessions |
| FDE hours live | 20 × 1.5 | 30 hours |
| Champion network | 1 per 25 users | 20 champions |
| Office hours | 2×/week × 8 weeks | 16 hours |
| Content creation | slides + videos | ~20 hours |
Enablement Touchpoints
| Channel | Cadence | Description |
|---|---|---|
| Live training | Week 0–2 | Role-based scenarios |
| Cheat sheet | Day 0 | Top 10 prompts + when to escalate |
| Office hours | 2× weekly | Drop-in debugging |
| In-app tips | First 3 sessions | Contextual nudges |
| Feedback form | Always | Feeds product backlog |
Data Model
| Entity | Key fields | Notes |
|---|---|---|
| learners | user_id, role, trained_at, champion_id | CRM-ish |
| sessions_attended | user_id, session_id | Compliance if needed |
| adoption_events | user_id, first_value_ts | Time-to-value |
High-level Design
Exec sponsor announce → Champion train-the-trainer (1 day) → Role cohorts (support vs ops curricula) → Day-1 cheat sheet + sandbox tenant → Weeks 1–4: office hours + tip emails → Week 2/4: adoption review → unblock UX issues → Graduate to customer L1 support ownership
Deep Dive: Teaching Prompt Patterns
- Teach task framing: goal, context, constraints, format
- Show failure demos (hallucinations) so users verify citations
- Role play escalations — when not to trust the bot
- Measure time-to-first-successful-citation, not attendance alone
Bottlenecks at Scale
- Manager no-shows — require team lead attendance
- Fear of job loss — address openly with sponsor
- One-size training — split by role workflows
Failure Modes & Monitoring
| KPI | Target | Alert |
|---|---|---|
| Training completion | ≥ 90% | Chase managers |
| WAU week 4 | ≥ 40% | Extra office hours |
| Time-to-first-value | < 3 days | Improve onboarding |
| Champion active | ≥ 80% | Replace inactive |