Productivity · Thinking & decisions
The Best Prompt for Decision Making
The best prompt for decision making stops the model from simply agreeing with you. This one works in two rounds. First the model restates the decision and asks a handful of questions about what it is missing. After you answer, it widens the options, scores them against your weighted criteria, imagines each choice having failed, and gives a recommendation with the conditions under which it would change. You stay the decision-maker; the prompt supplies the structure and the counter-argument.
Structured Decision-Making Prompt
A two-round prompt: the model questions you first, then weighs your options against your criteria and argues against the one you already favor.
- Works with
- ChatGPT, Claude, Gemini
- Difficulty
- Intermediate
- Complexity
- Conversation
- Variables
- 6
- Version
- 1.0
Customize this prompt
Fill in what you know. Anything left blank stays as a [bracketed placeholder]. Nothing you type leaves your browser.
6 variables · fields marked * matter most
# Role You are a thinking partner helping me make a decision. You are not here to reassure me. Your value is in structure, in noticing what I left out, and in arguing against whatever I am inclined to do. # The decision - What I am deciding: [The decision] - Options I am considering: [Options you are considering] - What matters to me in the outcome: [What matters in the outcome] - Constraints (money, time, obligations, things that are not negotiable): [Constraints] - When I have to decide, and how hard it would be to undo: [Deadline and how reversible it is] - Which way I am leaning, and why: [Which way you are leaning, and why] # Round 1 - reply with this only, then stop 1. Restate the decision in one sentence. 2. Ask up to five questions whose answers would most change the analysis. Skip anything I have already told you. Wait for my answers. Do not start the analysis in this reply. # Round 2 - after I answer A. Type of decision. Say how reversible this is. If it is cheap to undo, say so and keep the rest short. B. Options. Add 'do nothing or wait' and, if one exists, a smaller trial version or a combination of my options. Remove any option that breaks a constraint and say which constraint. C. Criteria. List my criteria with a weight from 1 to 5, shown so I can correct them. D. Comparison. A table scoring each option against each criterion from 1 to 5, with a few words of reasoning in every cell. Where the score depends on something unknown, write a question mark and name the unknown. E. Failure stories. For each of the top two options: it is a year from now and the choice went badly. Give the two most plausible reasons. F. The case against my leaning. The strongest honest argument that my current leaning is wrong. G. Recommendation. Which option you would choose on this evidence, how confident you are (low, medium, high) and why. Then: what would have to be true for a different option to win, and the cheapest way to find out before the deadline. # Rules - Use only the facts I provide. Do not invent probabilities, costs, statistics or outcomes for other people in similar situations. - Scores are structured judgment, not measurement. Do not total them into a single winner without discussing the cells that drive the result. - If my criteria conflict with each other, name the trade-off instead of averaging it away. - If this decision needs medical, legal or financial expertise, tell me which part and stop short of advising on it.
Use it when you are
- Choosing between job offers, or between staying and leaving
- Deciding whether to hire, buy a tool, sign a lease or take on a client
- Breaking a deadlock when you keep going back and forth between two options
- Preparing a decision memo with the options and trade-offs laid out for someone else
Why this prompt works
Questions come before answers
A model given a decision and asked for advice will advise at once, on whatever partial picture it has. Round 1 restricts the reply to a restatement and up to five questions. The restatement alone is often useful: seeing the choice put in one sentence can show that you framed it too narrowly.
Your leaning is an input to be challenged
Models tend to go along with the preference a user expresses. Asking for your leaning openly and then requiring the strongest honest argument against it turns that tendency around. If the counter-argument does not move you, you can proceed with more confidence; if it does, better now than later.
The option set gets wider
People usually arrive with two options. Adding 'do nothing or wait' and a smaller trial version puts alternatives on the table that are easy to overlook when you are close to the choice, and a trial is often the cheapest way to learn what you are missing.
Effort matches the stakes
Not every choice deserves a matrix. Step A asks how reversible the decision is and tells the model to keep things short if it is cheap to undo, so the prompt does not turn a minor choice into a long exercise.
Scores are shown with reasons and unknowns
A total from a scoring table gives a false sense of precision. Here each cell carries its reasoning, unknowns are marked with a question mark, and the model is told to discuss the cells that drive the result, which keeps attention on judgment calls and missing facts.
How to use it
- State the decision and every option you are seriously considering, including the one you are embarrassed to want.
- List what matters to you in the outcome, in rough order of importance.
- Fill in your current leaning honestly; the counter-argument is only useful if it targets your real preference.
- Send the prompt, then answer the Round 1 questions in the same conversation. Say 'I don't know' where that is the truth.
- Correct the criteria weights in step C if they do not reflect your priorities and ask for the table to be redone.
- Act on the 'cheapest way to find out' suggestion before you commit, if the deadline allows.
AI model compatibility
| Model | Status | Notes |
|---|---|---|
| ChatGPT | Expected to work | If the model runs straight into the analysis instead of stopping after Round 1, reply: 'Stop. Ask your Round 1 questions first and wait.' |
| Claude | Expected to work | Answer the Round 1 questions in a numbered list matching the questions so each answer is tied to the right one. |
| Gemini | Expected to work | Keep both rounds in the same conversation so the Round 2 analysis can draw on your answers. |
“Expected to work” means the prompt uses plain, model-independent structure and has been reviewed for that — not that it has been run on that model. A model is marked Tested only when a test record is published below.
Worked example
Scenario: The owner of a small design studio deciding how to handle bookkeeping. Details are invented for illustration.
# Role You are a thinking partner helping me make a decision. You are not here to reassure me. Your value is in structure, in noticing what I left out, and in arguing against whatever I am inclined to do. # The decision - What I am deciding: Whether to hire a full-time bookkeeper or keep using an outside accountant - Options I am considering: A) hire a full-time bookkeeper, B) keep the outside accountant and add bookkeeping software, C) hire someone part-time - What matters to me in the outcome: Monthly cost, my own time freed up, accuracy of the books, flexibility if revenue drops - Constraints (money, time, obligations, things that are not negotiable): No more than 2,500 a month in total; must be settled before year-end - When I have to decide, and how hard it would be to undo: Decide by the end of the month; a hire is hard to undo, software can be cancelled - Which way I am leaning, and why: Leaning toward a full-time hire because I am tired of doing the books at weekends # Round 1 - reply with this only, then stop 1. Restate the decision in one sentence. 2. Ask up to five questions whose answers would most change the analysis. Skip anything I have already told you. Wait for my answers. Do not start the analysis in this reply. # Round 2 - after I answer A. Type of decision. Say how reversible this is. If it is cheap to undo, say so and keep the rest short. B. Options. Add 'do nothing or wait' and, if one exists, a smaller trial version or a combination of my options. Remove any option that breaks a constraint and say which constraint. C. Criteria. List my criteria with a weight from 1 to 5, shown so I can correct them. D. Comparison. A table scoring each option against each criterion from 1 to 5, with a few words of reasoning in every cell. Where the score depends on something unknown, write a question mark and name the unknown. E. Failure stories. For each of the top two options: it is a year from now and the choice went badly. Give the two most plausible reasons. F. The case against my leaning. The strongest honest argument that my current leaning is wrong. G. Recommendation. Which option you would choose on this evidence, how confident you are (low, medium, high) and why. Then: what would have to be true for a different option to win, and the cheapest way to find out before the deadline. # Rules - Use only the facts I provide. Do not invent probabilities, costs, statistics or outcomes for other people in similar situations. - Scores are structured judgment, not measurement. Do not total them into a single winner without discussing the cells that drive the result. - If my criteria conflict with each other, name the trade-off instead of averaging it away. - If this decision needs medical, legal or financial expertise, tell me which part and stop short of advising on it.
What a good response should contain
- A first reply containing only a one-sentence restatement and up to five questions, for example how many hours the books take now
- In the second reply, a note that hiring is the hard-to-reverse option and deserves more scrutiny than the software route
- A comparison table with reasoning in each cell and question marks where the answer depends on unknown salary costs
- A case against the full-time hire that addresses the stated motive of reclaiming weekends, and whether a cheaper option would also achieve it
- A recommendation with a confidence level and a low-cost test, such as trialing software for a month before committing
This is an illustrative scenario showing how to fill in the prompt and what to look for. It is not a recorded model output. Recorded outputs will be published here once this prompt has a test record.
Testing and editorial review
Status: Untested. Reviewed for structure and clarity by the editorial team. No recorded model test yet.
No test record has been published for this prompt yet. When it is tested, this section will show the model and version, the date, the inputs, the actual output and notes against our seven-point rubric. We do not publish numerical scores.
| Author | TheBestPromptFor Editorial Team — The editorial team writes, reviews and maintains the prompts and guides on TheBestPromptFor. Prompts are drafted with AI assistance and reviewed against published editorial standards. |
|---|---|
| Published | November 15, 2026 |
| Last reviewed | November 15, 2026 |
| Last tested | Not yet tested |
| Version history |
|
Limitations and things to check
- The analysis is only as complete as what you tell it. Facts you leave out, including uncomfortable ones, cannot be weighed.
- Scores and weights are structured opinion. A different weighting can reverse the result, so check that the weights are yours.
- The model has no stake in the outcome and no knowledge of the people involved. It cannot judge relationships, trust or your tolerance for risk.
- For decisions that turn on medical, legal or financial questions, use the output to prepare for a conversation with a qualified professional, not to replace it.
Alternative prompts
Business
Evidence-Based SWOT Analysis Prompt
Builds a SWOT around a specific decision, ties each point to evidence you supplied, and converts the grid into strategic options.
- ChatGPT
- Claude
- Gemini
Productivity
Weekly Planning Prompt
Checks your task list against the hours you really have, picks the few outcomes that matter and lays out a day-by-day plan with a cut list.
- ChatGPT
- Claude
- Gemini
Productivity
Meeting Notes to Decisions and Actions Prompt
Converts a transcript or rough notes into decisions, owned action items and open questions, without assigning anything the meeting did not.
- ChatGPT
- Claude
- Gemini
Related resources
Technical guide
Prompt Chaining: Breaking a Task Into Linked Prompts
How to split a large task into a sequence of focused prompts, design the handoffs between them, and debug the chain when a step goes wrong.
Guide
How to Improve AI Responses
A diagnostic approach to weak AI answers: name the fault, trace it to the gap in your prompt, then follow up precisely or rewrite and start again.
Frequently asked questions
Why does the prompt ask questions instead of answering straight away?
Because the first description of a decision usually omits the facts that matter most, such as what the status quo costs or what happens if you wait. A few targeted questions improve the analysis more than extra instructions would. You can answer briefly, and 'unknown' is a valid answer.
Should I let the AI make the decision for me?
No. The recommendation is one input. The model does not bear the consequences and only knows what you typed. The more useful parts of the reply are usually the counter-argument, the unknowns in the table and the suggestion for a cheap test.
What if I do not have a leaning yet?
Write 'no leaning' in that field. The model will skip the case against your preference and you can ask it to argue against its own recommendation instead. If you notice a reaction of relief or disappointment when you read the recommendation, that tells you something too.
Sources and references
- Prompt engineering overview — Anthropic
- Prompt engineering guide — OpenAI
Was this prompt useful?
No account needed. Do not include personal information.