CONCIERGE AI PILOT WORKSHEET

Save a copy and agree the relevant details with your pilot participant. The worksheet organizes a test; it is not a substitute for suitable commercial or data-handling terms.

1. CUSTOMER RESULT
Customer group and task:
Recognizable output:
What the customer will do with it:
What would make it unusable:
What we need to learn:

2. PILOT CHARTER
Start / finish / review date:
Number and timing of deliveries:
Required inputs and access:
Work performed manually:
Work assisted by AI:
Human review and approval owner:
Scope, exclusions, and change requests:
Price or explicit free arrangement:
Relevant cancellation / refund / continuation terms:
Data permissions, access, retention, and deletion:
Exception and escalation process:

3. BASELINE
Current workflow and boundaries measured:
Frequency and people involved:
Active effort and elapsed time:
Quality problems and consequences:
Which figures are measured and which are estimates:
Agreed success conditions:

4. DELIVERY LOG (repeat for each request)
Request code and date:
Input complete? If not, what was missing?
Human minutes, including review and chasing inputs:
Tool usage / direct cost:
Corrections, exceptions, and unresolved questions:
Quality check and reviewer:
Customer action after receiving the result:
Customer feedback and next request:

5. DECISION
Evidence of usefulness:
Evidence of willingness to pay:
Evidence about repeatability:
What this manual test cannot establish about software:
One bounded step that may be automated:
Representative normal, messy, and declined cases for evaluation:
Fallback owner and stop condition:
Decision: continue / narrow / change model / stop
Next owner and review date:

Keep human-delivered service evidence separate from autonomous software performance. Do not hide manual work or infer product-market fit from a handful of pilots.

Guide: https://startupcoach.cc/journal/test-ai-startup-concierge-mvp
StartupCoach editorial | October 2026