AI Decision Matrix

How a weighted decision matrix reaches a verdict?

A decision matrix separates two things people normally mix together: how much a criterion matters to you, and how well each option performs on it. Once those are separate numbers, the arithmetic is trivial and the argument moves to where it belongs — the weights.

Each criterion carries a weight, and each option receives a score against that criterion. The tool multiplies score by weight, sums across criteria, and normalises the totals so the options can be compared on a single scale regardless of how many criteria you added.

The interesting output is rarely the winner. It is the margin. When two options finish within a few percent of each other, the matrix is telling you the decision is close enough that speed, reversibility or gut preference are legitimate tie-breakers. A wide margin tells you the opposite: you already know the answer and are looking for permission.

How do you use the AI Decision Matrix?

1. 1.List every option you are genuinely willing to choose, including the option of doing nothing.
2. 2.Write criteria as outcomes you care about, not features. 'Time to first result' beats 'has an API'.
3. 3.Set weights before you score anything, so the weights are not reverse-engineered to produce your favourite answer.
4. 4.Score each option honestly, then look at the margin. If the top two are within five percent, pick the more reversible one and move on.

What can this tool not tell you?

• Garbage weights produce a confident wrong answer. The matrix formalises your judgement, it does not supply any.
• It assumes criteria are independent. When two criteria measure nearly the same thing, that factor is silently double-counted.
• It cannot model thresholds. If an option fails a hard requirement, remove it rather than letting a high score elsewhere rescue it.

Why weights argue better than gut feel?

Most disagreements about a decision are actually disagreements about priorities dressed up as disagreements about facts. Two colleagues can agree completely on how each option performs and still reach opposite conclusions because one silently weights cost at fifty percent and the other weights it at ten. A matrix forces that hidden number into the open, which usually shortens the argument by an order of magnitude because everyone is finally debating the same thing.

The tool is deliberately blunt about normalisation. Raw weighted totals are hard to compare when people use different scoring ranges for different criteria, so results are rescaled onto a common band before ranking. This stops a criterion scored out of ten from silently dominating one scored out of five, an error that is common in matrices built by hand in a spreadsheet without anyone checking the scales line up.

A well-run matrix session usually surfaces one uncomfortable finding: an option everyone privately favours performs worse once the weights are written down honestly. That moment is the actual value of the exercise. The spreadsheet is not there to replace judgement, it is there to catch the gap between what you say matters and what you have quietly already decided.

The most useful thing a matrix produces is often not the winner but the argument it starts. When two people who have been circling a decision for a week each fill in weights separately, the disagreement almost always turns out to be about the weights rather than the scores — they agree the commute is forty minutes, they disagree about how much forty minutes should count. That is a far more tractable conversation than 'which job is better', because it is about values that can be stated, and once stated it usually resolves in minutes. A matrix that surfaces a weight disagreement has already earned its keep even if nobody looks at the totals afterwards.

What do worked examples look like?

What do people ask most about this tool?

Which related tools should you try next?