Value Hypothesis Worksheet

Use this short template to craft a testable hypothesis that ties AI to a measurable outcome.

1. Pilot name: ________________________
2. Business outcome we care about (in plain language):

   Example: Reduce average manual review time for invoices.

   Answer: ________________________

3. Value hypothesis (one sentence):

   Format: If we apply [model/automation] to [process], then [metric] will change by [target] within [timeframe], without increasing [negative outcome].

   Answer: ________________________

4. Primary metric: ________________________ (How will you measure value?)
5. Guardrail metrics (min 1):

   Examples: error rate, percent of cases requiring human override, false positive rate.

   Answer: ________________________

6. Minimum success threshold to consider scaling:

   Example: 20% improvement on primary metric and < 3% increase in error rate.

   Answer: ________________________

7. Primary risks & mitigations:

   List the key risks and a short mitigation for each.

   Answer: ________________________

8. Key stakeholders & roles:

   Sponsor, owner, data steward, domain reviewer.

   Answer: ________________________

Keep this worksheet with your pilot record and review it at every checkpoint.