Step 1 / 5 · 15 min
Make the alternative explicit before celebrating the AI demo.Price the current workflow
- Measure current time, quality, and failure cost.
- Name the existing tool or human workaround.
Shape · Core skill lab
AI Feature Viability Brief
Decide whether AI creates a defensible workflow advantage after trust, latency, and cost are counted.
- Decision
- Should this workflow use AI now—and where must human authority remain?
- Timebox
- 90 minutes
- Output
- AI feature viability brief
Bring to the bench
- A specific user workflow
- Its current non-AI baseline
- Rough volume, latency, and value assumptions
Your work saves in this browser.
Field tools
Use the instrument, not a blank page.
Copy these into your interview, agent, review, or working document. They are specific to this repetition.
template
Outcome economics worksheet
Compare successful outcomes, not API calls.
CURRENT COST / SUCCESS = labour + tools + rework + expected failure loss AI COST / SUCCESS = model + retrieval/tools + human review + support + expected failure loss VOLUME / month: SUCCESS RATE: GROSS VALUE / success: MAX AFFORDABLE COST / success: KILL THRESHOLD:
checklist
Human authority map
Every high-consequence AI action needs an explicit answer.
Who can inspect the evidence? Who approves before action? Who can edit or override? Who handles uncertainty? Who receives an appeal? What happens when the model or dependency is unavailable? Which action is the model forbidden to take?
Calibrate judgment
Compare the evidence, not the polish.
Assistive by design
A contract handoff workflow contains high-consequence claims.
Use AI to extract candidate constraints with source links, never to approve the brief. Ship only after ≥95% recall and zero uncited claims on the evaluation set. The account lead retains approval and a manual path. Kill if review remains above 30 minutes after 50 briefs.
Why it works: The model role, human authority, launch threshold, and kill condition match the failure cost.
Model access as strategy
Add GPT-5 to automate onboarding. It will be faster and smarter than competitors, and we can add guardrails later.
Why it fails: The workflow, failure cost, authority, economics, measurable advantage, and durability are undefined.
Review → revise → repeat
The artifact is the beginning of the rep.
Check only standards your current artifact actually meets. Then record one consequential revision before exporting it.