Reasons to move forward
Continue when these are mostly true.
- The task is narrow and repeatable.
- Source information is appropriate to use.
- Mistakes can be detected before harm occurs.
- A named person owns the final decision.
Decision guide
Should this workflow use AI?
Maybe—but begin with the task, the risk, and the person accountable for the result. If those are unclear, adding AI usually adds uncertainty rather than removing it.
Four questions
Pause here before choosing.
You do not need perfect answers. The point is to make the uncertainty visible.
What is the task?
Describe the input, the decision or output, and the person who uses it.
What could go wrong?
Consider privacy, accuracy, fairness, confidentiality, safety, and the cost of a convincing mistake.
Can a person review the result well?
A human review step only works when the reviewer has enough time, context, and authority to reject the output.
Would a simpler change help first?
A clearer template, better source information, or a smaller handoff may solve the problem without AI.
Reasons to pause
Wait when these remain unresolved.
- The purpose keeps changing.
- Sensitive information has no approved handling path.
- The result will be trusted without meaningful review.
- The organization cannot explain or correct the process.
A simple starting point
Want help applying this guide to a real situation?
Share the question and enough non-sensitive context to understand it.