AI Meeting Notes: Check That the Decision Is Still Yours

A small studio agrees to try a new newsletter format with one group of readers. Nobody commits to changing the main edition. An AI meeting summary lands an hour later: the team agreed to roll out the new format next week. The wording sounds familiar. A colleague replies that a rollout makes sense because the new layout is easier to read. That may be a good argument for rolling it out. It is also an argument for a decision the studio did not make.

The moment to pay attention is when the new version starts attracting reasons of its own. Once a team is discussing whether a rollout makes sense, it can lose track of a simpler question: was there ever a rollout decision? Research on choice blindness gives us a reason to preserve the answer somewhere other than the summary.

How people explain a choice they did not make

In a 2005 study published in Science, Petter Johansson and colleagues asked participants to choose the more attractive of two faces. On some trials, the researchers covertly presented the face the participant had passed over as if it were the one they had chosen. Participants sometimes failed to spot the switch and gave reasons for a preference they had not expressed. The researchers called the effect choice blindness.

Lars Hall and colleagues tested the idea in a different setting in a 2010 study in Cognition. At a supermarket, people compared the taste of jams or the smell of teas. The researchers sometimes swapped the selected sample for the alternative when presenting it again. Here too, some people missed the substitution. The studies used deliberately engineered switches in particular tasks. They do not show how often AI meeting notes change decisions or how often teams would catch a change.

The editorial inference is narrower: the ability to explain an outcome does not establish that it is the outcome someone originally chose. A colleague who can say why a rollout sounds sensible may be thinking well about the version now in front of them. They may still need to look at the original agreement before endorsing the notes.

Find the word that changed the commitment

In the studio example, the new format did not disappear from the record. The affected detail was the scope of the decision. A trial with one group became a launch to all readers. Other shifts are equally small on the page: a person who offered to explore a supplier becomes the owner of a purchase; a deadline someone proposed becomes a date everyone accepted; a condition about checking accessibility disappears from an otherwise accurate action item.

A general request to review the notes may draw attention to names, dates and spelling. Instead, read commitments as separate sentences. What is going to happen? Who agreed to do it? How far does it extend? What has to happen first? If the answer is absent from the original record, say so. A short statement that the decision is still open is more useful than an elegant paragraph that settles it by accident.

A similar distinction appears in our article on what a quick AI reply can tell you. A smooth exchange gives you a particular kind of experience, but the speed of a response and the accuracy of its understanding are separate questions. Here the thing to check is whether a coherent note preserves the decision that people actually made.

Keep the record short enough to use

At the end of a meeting, write one plain sentence for each decision with consequences outside the room. In this example, the sentence could say that the studio will test the new layout with one subscriber group and will decide whether to expand only after reviewing the trial. It can sit in the same shared document as the AI notes. Give the sentence a stable place and let the meeting owner correct it before other work depends on it.

You do not need to capture every passing suggestion. For a small team, it may be enough to record decisions that change a delivery date, assign responsibility or affect customers. Let a summary organise the rest. If no decision was made, write that down too. That way, a suggested action can stay visible without silently becoming an instruction.

• Capture the agreement before polishing the notes. Record scope, owner and any conditions in plain language.
• Compare the decisions line by line. Ask which words in the summary change what someone will do.
• Allow a new decision to be new. If the revised idea is better, record it as a fresh decision with the people who need to agree.
• Keep the burden proportionate. Check consequential commitments carefully and give routine discussion a lighter review.

For a solo creator, the same habit works without a meeting. Before asking an assistant to turn rough ideas into a plan, write the one or two choices you have already made. After reading the plan, compare those choices with the recommendations. You can still change your mind. The value of the comparison is that you know when you are doing it.

Sources

• Johansson and colleagues, study of mismatches between choices and outcomes, Science (2005).
• Hall and colleagues, choice blindness in jam and tea preferences, Cognition (2010).