Consulting: Scope Shaped by Verification, Not Features

Five months of AI-assisted consulting has rewritten my scoping approach. The old way was to list features and estimate hours. The new way is to list verifications and estimate review time. That shift comes from a painful lesson: AI can generate faster than I can verify, and the unverified work isn't work.

The Old Scope Shape

Old scope documents listed what I'd build. Five endpoints, three integrations, one dashboard. The client signed off. I built. We argued later about what 'done' meant.

The New Scope Shape

New scope documents list what I'll verify. For each feature: what specifically will I test, what output will I show the client, what signs do we agree mean it works. The hours estimate is driven by that verification list.

Feature: subscriber API
Verification:
  - 3 automated tests (happy path, validation, dedup)
  - live POST in staging with screenshot
  - admin dashboard shows new row
Review time estimate: 2 hours

That's six lines of scope for one feature. The client knows exactly what 'done' looks like. I know exactly how much review budget I'm using.

Why This Matters More With AI

AI-assisted generation is 10x faster than verification. If I scope by generation hours, I undersell. The work that actually takes time is the testing, the review, the demo, the client conversation. Scoping by verification prices the thing that's actually scarce.

The clients I've scoped this way haven't pushed back. They prefer knowing how they'll see the work is done.