Evidence enters through human review, not blind auto-acceptance.
1. Scan
Start outside-in. Before interviews begin, dated public evidence from customers, sales channels, competitors, suppliers, regulators, hiring, investors, and industry sources can create a focused discovery agenda. For consumer businesses, aggregate reviews and platform policies can serve the same purpose. Read the full demand intelligence method.
2. Source
Add company websites, documents, spreadsheets, annual reports, interview notes, and optional employee-approved workflow summaries. Machine-drafted facts remain suggestions until a person reviews them.
3. Evidence
Connect workflows, problems, assets, limits, and financial signals before ranking opportunities. Keep sources attached so reviewers can see what is known and what is still missing outside the app.
4. Review
Compare the evidence with 101 opportunity patterns. Show which ideas the evidence supports, test value and systems readiness, explain why each one ranks where it does, and separate fund-now ideas from preparation work and rejected ideas.
5. Prove
Plan adoption by ordering foundations, owners, controls, readiness questions, and rollout. For suitable agents, run an observe-only simulation on approved historical or clearly labeled synthetic cases. Fixed checks return ready, needs work, or unsafe instead of allowing an AI model to approve its own output.
6. Build
After the required checks pass, prepare the agent design, supervised pilot plan, starter code, and production handoff package. Use Case Foundry prepares the package; the customer's technical owner supplies credentials, verifies integrations, and decides what goes live.