Skip to main content

Validating an AI Use Case

Purpose​

A scenario for running an experiment on an AI initiative: verifying feasibility, data, and initial value in order to move to "Prototype" (Gate 2).

Core Ideas​

  • The experiment is limited in time and scope: the goal is to prove or disprove the hypothesis, not to build a complete solution.
  • What is verified: data quality and availability, the suitability of the model/approach, measurable metrics (accuracy, business proxies).
  • The result is documented as a report for Gate 2: a recommendation to "go to prototype," "refine the experiment," or "close."

How It Works​

  1. Experiment plan: goals, success metrics, required data, timeframe (for example, 2–4 weeks).
  2. Data access: alignment with data owners and governance; the minimum necessary dataset for the experiment.
  3. Execution: building/testing the model, measuring metrics, recording constraints and risks.
  4. Documentation: a report with conclusions, an updated impact assessment, and a recommendation.
  5. Gate 2: decision to move to "Prototype," iterate on the experiment, or close the initiative.

After a successful Gate 2 comes the transition to the build-ai-prototype playbook.