Many people use AI with questions like, "Should I choose A or B?" The problem is, such questions often yield answers that sound confident but may not reflect the actual situation. AI might choose the popular, cheaper, or more comprehensive option without understanding what is most important to you.
For decisions with multiple considerations, use AI as a comparison tool, not as a substitute for judgment. One practical way is to ask AI to create a decision matrix. Simply put, each option is evaluated based on criteria you define, and then AI explains the advantages, disadvantages, and risks of each.
Why is the question "which is best?" often unhelpful?
The word "best" almost always depends on context. The best laptop for a video editor may not be the best for a student. The best note-taking app for one person might feel cumbersome for another. Even a low price doesn't always mean savings if the costs of data migration, training, or additional subscriptions are high.
When criteria are not specified, AI has to guess your priorities. This is where answers can go awry. AI may provide recommendations that make sense generally, but do not fit your specific needs, budget, skill level, or risk tolerance.
A decision matrix helps turn vague questions into verifiable comparisons. You not only receive a conclusion but also see the reasoning behind that conclusion.
Basic structure of a decision matrix prompt
A good prompt doesn't need to be long and complicated. The important thing is that AI knows five things:
- What decision is being made.
- What options are to be compared.
- What evaluation criteria are important.
- The weight or level of importance of each criterion.
- The form of the results that are easy to verify.
You can use the following template:
I am choosing [type of product/service/strategy] between [option A], [option B], and [option C].My main goal: [describe the desired outcome].Rate each option based on the following criteria on a scale of 1-5:- [Criterion 1], weight [percentage or level of importance]- [Criterion 2], weight [percentage or level of importance]- [Criterion 3], weight [percentage or level of importance]Create a score table, explain the assumptions used, display the weaknesses of each option, and provide recommendations based on my priorities. If there is missing information, mention it before drawing conclusions.The weights do not always have to be numbers. You can use categories like "very important," "important," and "additional." However, numbers help AI make more consistent comparisons.
Example: choosing a project management app
For instance, you are comparing three apps for a small team. Don't just write, "Which is the best?" Try providing context like this:
I manage a team of six and want to choose a project management app for client work. Compare Trello, Asana, and ClickUp based on ease of learning, deadline management capabilities, comment collaboration, integration, and cost. Ease of learning and cost are very important because not all team members are technical. Provide scores from 1-5, explain the reasons for each score, and highlight parts that I need to verify myself from the official sites of each service.
This prompt is more useful because AI is not forced to answer based on popularity. It must consider team size, user capabilities, and budget constraints. You also request parts that need verification, so the results are not taken as final truths.
Add scenarios so recommendations are not too general
Decisions often change when conditions change. Therefore, after obtaining the initial table, ask AI to perform scenario analysis.
Re-evaluate the recommendations under three conditions: a 30 percent budget cut, an increase in team members to 20, and the project requiring more automation. Show which options changed positions and explain why.
This step is useful because the best choice today may not remain suitable six months from now. You can also find out if AI's recommendations rely too heavily on one criterion alone.
Use AI to find weaknesses, not just to justify choices
After AI provides recommendations, don't immediately ask for a summary. Ask it to critique the results from an opposing viewpoint.
Act as a critic of the above recommendations. Identify the three weakest assumptions, two risks I might have overlooked, and conditions that could make the second option better than the first.
This prompt helps reduce the human tendency to seek justification. If you favored one option from the start, AI can be used as a checker, not just a machine that compiles reasons to support that choice.
Differentiate data, assumptions, and judgments
The results of comparisons will be more trustworthy if each part is labeled. Ask AI to separate:
- Data: information provided or known from specific sources.
- Assumptions: things that have not been confirmed and are used to complete the analysis.
- Judgments: interpretations based on the priorities you provide.
Example instruction:
Separate the answer into three parts: available facts, assumptions that still need verification, and analysis or opinions based on my criteria. Do not disguise assumptions as facts.
This is especially important when comparing prices, features, service policies, security, or product specifications. Such information can change, so it is best to verify from official sources before making a final decision.
Errors to avoid
- Including too many criteria. If everything is considered important, AI will struggle to distinguish between primary and additional priorities.
- Using scores without definitions. Explain what scores 1 and 5 mean so that evaluations are not too subjective.
- Ignoring migration costs. Data migration, training, and adaptation time often exceed subscription prices.
- Accepting recommendations without verification. AI can be wrong, especially regarding product and service information that is frequently updated.
- Mixing small and large decisions. For decisions involving health, legal matters, or large sums of money, use AI only to formulate questions and analysis frameworks.
What you can do now
Choose one real decision you are facing this week. Write down three to five options, then define up to five criteria. Mark which are truly important and which are just bonuses. After that, use the decision matrix prompt to create an initial comparison.
Don't stop at the option with the highest score. Also ask: "What information is most likely to change this outcome?" This question transforms AI from a mere answer provider into a tool for clarifying thought processes.
Ultimately, a decision matrix is not a fate-deciding machine. It merely makes considerations that are usually overlooked more visible. The decision remains in your hands, but at least you can explain why you chose one option, what risks you accepted, and what needs to be verified before taking action.
– Rio Yotto @rioyotto