Case 01·Intermediate·~25 min
The Disappointing AI Agent
A $300K deployment is live, but automation is stuck at 8%.
Phase 1 · Case
Your startup sells an enterprise AI customer-support platform. A customer has signed a $300,000/year contract, and you are the FDE responsible for deployment.
Two weeks after launch, only 8% of tickets are fully automated. The customer's VP of Support is unhappy: “We expected something closer to 40% automation.”
What you know
- FAQ-style tickets perform well.
- Most tickets require order information stored in a legacy internal system.
- The AI cannot currently access that system.
- Security may allow controlled read access but opposes autonomous write access.
- Agents frequently re-check AI-generated answers and do not fully trust the system.
- Average handling time is 15 minutes.
- Ticket volume is expected to grow approximately 30% next year.
- Fully loaded support-agent cost is unknown.
- Sales presented 40% automation, but it is unclear whether that was the business goal or a proxy for efficiency.
Phase 2 · Your Analysis
Reason from customer outcome to workflow, measurement, economic impact, and field leverage. Label uncertainty instead of inventing facts.
Sign-in is required when you submit so your progress can be saved.