In How to Use ChatGPT for Better Everyday Decisions I described a five-step method: frame, generate alternatives, define what matters, compare and decide. For bigger decisions, step four deserves a proper tool. That tool is the weighted decision matrix, and ChatGPT makes building one fast.
A decision matrix is just a table. Options go down the side, criteria go across the top, each criterion gets a weight, and each option gets a score. Multiply, add and you have a ranking. The value isn’t the arithmetic. It’s that the matrix forces you to say out loud what matters and how much.
When a matrix is worth it
Use one when:
- you have three or more real options
- the options win on different things (one is cheaper, another is more reliable)
- the decision is expensive or hard to reverse
- other people need to understand or agree with the choice
Skip it for decisions you could reverse tomorrow at no cost.
Step 1: List the options
Give ChatGPT your framed decision and ask for options. Keep the final list between three and seven. More than that and the scoring becomes noise.
Prompt: “I’m choosing a project management tool for a 12-person consulting team. We need client-visible boards, time tracking and a reasonable price. List six realistic options, one line each.”
Cut the list down yourself. Add any option you already know is in the running.
Step 2: Choose the criteria
Ask for criteria, then edit hard. Five or six is plenty.
Prompt: “Suggest criteria for this decision. For each, explain in one sentence why it matters for a team like ours.”
Make every criterion something you could actually score. “Ease of use” is fine if you’ll trial the tools. “Future-proof” is not, unless you define what it means.
Step 3: Weight the criteria
Give the criteria weights that add up to 100. This is the most important step, and it’s yours, not the AI’s.
A useful trick: write your own weights first, then ask ChatGPT to propose weights based on everything you’ve told it. Where the two differ a lot, stop and think. Either you’ve told the AI something misleading, or you’ve discovered that you care about something more than you assumed.
Step 4: Score each option
Score each option on each criterion from 1 to 5. Ask ChatGPT for a first pass, with a short reason for every score:
Prompt: “Score each option 1 to 5 on each criterion. Give a one-line reason for every score, and mark any score that depends on information you’re not sure about.”
Then check the marked scores. Pricing pages, feature lists and trial accounts beat model memory every time. Change any score you disagree with. You’re the one who has to live with the result.
Step 5: Calculate and read the result
Ask ChatGPT to produce the table with weighted totals, or paste it into a spreadsheet. A sample layout:
- Criteria and weights: Client boards (30), Time tracking (25), Price (20), Ease of use (15), Integrations (10)
- Option A: 5, 3, 3, 4, 4 → weighted total 3.9
- Option B: 4, 5, 4, 3, 3 → weighted total 4.0
- Option C: 3, 4, 5, 5, 2 → weighted total 3.85
Close totals like these are common, and they tell you something: the options are similar on paper, so the decision will turn on one or two criteria. Look at where they differ most.
Step 6: Stress-test it
Before deciding, run two checks:
- Sensitivity: “If I change the weight of price from 20 to 30, does the winner change?” If a small change flips the result, the decision is close, and a trial or a conversation with a real user is worth more than more analysis.
- Gut check: If the winner makes you uneasy, ask yourself why. Unease often points at a missing criterion. Add it and rescore.
Step 7: Decide and keep the matrix
Choose, and save the matrix with a two-sentence note on why. If someone asks later why you picked Option B, you’ll have the answer. If circumstances change, you can update the weights instead of starting over.
Keep the human in charge
ChatGPT is excellent at producing options, criteria and first-pass scores quickly. It’s less reliable on current facts, and it has no stake in the outcome. The matrix works because it makes the AI’s contributions visible and checkable, and keeps the weighting, verification and final call with you. For the traps to watch for, see Six Mistakes to Avoid When Using AI for Decisions. For a real, slightly unusual example of this method in action, see how I used it to compare eight LA Phil holiday concerts.