Use this when
The work is real, but the starting point is not.
- Your team can name many AI ideas but cannot agree where to start.
- The proposed work spans several people, files, or tools.
- You need a reversible first test with a clear baseline.
Run the guide
Move from source material to a reviewed result.
- 01
List recurring work
Use real verbs such as prepare, reconcile, summarize, compare, draft, route, review, update, or report.
- 02
Score value
Rate frequency, time drain, delay or error cost, and handoffs from zero to three.
- 03
Score buildability
Rate repeatability, input readiness, safe testability, and measurability from zero to three.
- 04
Apply the risk gate
Name the human reviewer, allowed data, and approval point for every sensitive or consequential step.
- 05
Set the delegation mode
Choose Automate for low-stakes reversible steps, Accelerate for judgment-heavy work with review, or Stay Human when accountability cannot be delegated.
Delegation map
Choose how much of the work AI should own.
Low stakes and reversible
Use for repeatable steps with clear inputs, observable results, and a safe recovery path.
AI drafts, a person decides
Use when speed helps but judgment, context, trust, or quality still needs a responsible reviewer.
Accountability cannot be delegated
Keep people in control when the work carries high consequences, sensitive relationships, or unclear authority.
Starter prompt
Use this with your real source material.
Help me choose the first AI workflow worth testing. Interview me about recurring work, value, inputs, risk, and measurement. Recommend one reversible first test and state the strongest reason not to choose it.
Keep control
Review the parts that matter.
- A high score never overrides sensitive-data, advice, send, money, permission, or compliance controls.
- Keep the first test offline and draft-only when the approval point is unclear.
- Test three real examples, including one awkward edge case, before scaling.