Quick Answer
A good decision prompt asks the model to compare options against criteria, explain tradeoffs, identify missing information, and separate recommendation from uncertainty.
Use this guide when
The reader wants AI help with decisions without getting oversimplified advice.
Working Method
The practical move is to make the model's job visible. Before you ask for the final output, define the important choices you do not want the model to guess.
- Name the decision and the options under consideration.
- List criteria in priority order.
- Ask for assumptions and information that would change the recommendation.
- Request a tradeoff table before a recommendation.
- End with a low-risk next step rather than a forced final answer.
Prompt Example
Too vague
Should we launch now or wait?More useful
Compare launching our beta next week versus waiting four weeks. Criteria in order: customer trust, learning speed, support load, and engineering risk. Use a tradeoff table, list assumptions, and recommend a low-risk next step rather than a final go/no-go decision.Common Pitfalls
- Asking for a decision without giving options or criteria.
- Letting the model optimize for the wrong priority.
- Ignoring missing information that could reverse the conclusion.
How to Judge the Answer
A better prompt is only useful if the answer becomes easier to evaluate. Before using the response, check whether it meets the standard you set.
- The answer shows tradeoffs rather than only a recommendation.
- Criteria are applied consistently.
- The next step reduces uncertainty.
FAQ
Should AI make the decision?
No. AI can structure tradeoffs and questions, but accountable humans should make consequential decisions.
How many options should I compare?
Two to four options are usually easier to evaluate clearly. More options may need a first pass to group or eliminate choices.
Sources
Selected references that informed this guide:
- Overview of prompting strategies Google Cloud
- AI Risk Management Framework NIST