Decision Compiler
Think through an AI decision.
Answer five questions about the work, the evidence, and the risks. Get a starting recommendation for what to test, what to measure, and where people stay in control.
- Five questions
- Transparent rules
- Runs in your browser
Your decision
Frame the product before the model.
Describe the consequence, evidence, learning speed, repetition, and change involved. Your recommendation updates as you answer.
What your answers imply
Each input shapes a part of the plan.
- DecisionRecommendAligned
- WorkflowDaily / TeamAligned
- ModelReviewable to verifyAligned
- GuardrailsMaterial error costReview
- FeedbackDaysAligned
What changed
Change any answer to see how the recommendation responds.
The same answers always return the same recommendation.
Recommended role for AI
Recommend
Let AI propose a next move, but make evidence and user judgment part of the product.
Why this recommendation
- Errors create meaningful cost, so user judgment must remain inside the operating loop.
- Skilled review is possible, so recommendations can be tested without hiding uncertainty.
- Delayed feedback requires disciplined reconciliation rather than proxy-only success.
- Start here
- Shadow 50–100 representative decisions, then expose ranked recommendations with evidence to a small expert cohort. Compare decision quality, elapsed time, and correction work against the current baseline.
- What to measure
- Representative task set anchored to the current human-workflow baseline
- Expert rubric, evidence grounding, and reviewer-disagreement analysis
- Delayed outcome reconciliation segmented by workflow and user behavior
- Weekly slice review across user, task, and failure type
- Safeguards
- Show supporting evidence and preserve explicit user approval
- Make reject, edit, and escalation paths first-class product actions
- What the team needs
- Agree on a shared review rubric, workflow owner, escalation path, and team-level adoption measure before rollout.
- Stop if
- Stop if recommendations fail to improve decision quality or time-to-decision after accounting for added review work.
The thinking behind it
Make the reasoning visible.
A model’s capability is one part of the decision. This tool looks at the conditions around it: the work, the people, and the consequences.
- Value: start with the work
Be clear about the decision or task you want to improve before choosing how AI should help.
- Evidence: make progress testable
Choose a first test and useful measures. Consider whether you can verify the output and learn from mistakes.
- Control: define the boundaries
Decide who approves an action, when a person steps in, and how the team can reverse a mistake.
This is a deterministic decision aid, not a readiness score or a substitute for discovery. The same answers always produce the same recommendation.
Working through a consequential AI decision with your team? Explore working sessions.