Back to Insights

The Meeting of the Bots

AI can produce an accurate record of a meeting where nothing was decided. The Decision Minute is the five-minute fix.

By Jerald Lee

In brief

Two assistants can summarize a meeting perfectly. That is exactly how a room can look decisive and decide nothing.

Two assistants can now attend a meeting, transcribe it, summarize it, extract the actions, and send the follow-up before the room has cleared. The summary will be accurate. It is also possible for the meeting to have decided nothing and for the record to look complete. Recording, summarizing, and following up are solved. Deciding is not, because a decision needs authority, a named owner, and a condition that would reverse it. A better record makes the gap harder to see, not easier. Five minutes at the end, with the bots off, closes it.

# The Meeting of the Bots

The summary arrives before the room has cleared.

It is good. Better than most human minutes. It captures the shape of the discussion, attributes the points correctly, lists the actions with names and dates. If you missed the meeting, you can read it in three minutes and know what happened.

What it cannot tell you is whether anything was decided.

What the bots are good at

Three things, and they are not small.

Recording. Every word, attributed, searchable. This is a genuine improvement on the era of one person taking notes while everyone else pretended to listen.

Summarizing. The shape of the conversation, the points raised, the positions taken. A good summary is faster than a good minute-taker and less likely to reflect one person's view of what mattered.

Following up. Actions extracted, owners named, deadlines proposed. The administrative tail of a meeting has largely been solved.

A lot of meeting time was spent on exactly this, and it is now cheap. The mistake would be to treat that as the whole job.

What they cannot do

Decide.

A summary can record that two options were discussed and no choice was made. It cannot record that the choice is now overdue, or that the person who has to make it is avoiding it, or that the room was waiting for someone to say the uncomfortable thing and nobody did.

It also cannot carry authority. A follow-up email from an assistant is a record. A decision from a room is a commitment. The two look similar in a transcript and are not the same thing.

And it cannot hold the pause. The three seconds before someone says the thing that changes the meeting. The bots transcribe the sentence, not the silence that made it possible.

The authority gap

Here is the pattern worth watching, and it is new enough that most leaders have not named it.

A meeting runs on bots. The summary is excellent. The actions are clear. Everyone feels the meeting was well run, because the record is good.

But the decision that needed a room did not happen, because the room was not really a room. It was a set of people producing content for a summary.

The better the record, the easier it is to miss that nothing was decided. A bad minute-taker at least left you unsure. A good summary leaves you confident.

That is the mechanism. The record used to be the weak signal, so an absence of decisions showed up as an absence of minutes. Now the minutes are perfect and the absence has nowhere to appear.

The Decision Minute

Five minutes at the end of a meeting, with the bots off. One page, four lines.

What did we decide? One sentence. If there is no sentence, nothing was decided, and the meeting says so.

Who owns it? A name, not a team.

What would make us reverse it? The condition that would reopen the decision. This is the line that prevents a decision from being relitigated every month.

When do we check it? A date.

The page goes in the same place as the summary, so the record shows both what was discussed and what was decided. The gap between the two is usually the most useful thing in the file.

The third line does more work than it looks like. Most decisions get reopened because nobody wrote down what would count as new information. Without that line, a decision is only as stable as the last person who disagreed with it.

Why the obvious objection is reasonable

Three objections come up, and all three are fair.

"This is more admin on top of a full calendar." It is five minutes, and it replaces the follow-up meeting where the same conversation happens again because nobody remembers what was agreed. The admin is already being paid. It is just being paid later, at a higher rate.

"Our meetings are fine." Possibly. The test is not how the meeting felt. It is whether you can name the last three decisions it produced, and who owns each one.

"The summary is for the people who were not there." True, and it is a good use. The summary is a communication tool. The decision minute is a commitment tool. One document cannot do both jobs, which is why the second one has to be written separately.

Why this is not nostalgia

This is not an argument for worse meetings, or for doing things by hand, or for a return to the days when one person's notes were the only record.

The bots are better at recording than we are. Let them record. The question is what the meeting is for.

If it is for producing a record, the bots have won and there is nothing left to discuss. If it is for making a decision that a group of people will actually commit to, then the part that matters is the part that does not transcribe.

There is a practical test. If the summary is deleted tomorrow, does anything change? If the answer is no, the meeting produced a document. If the answer is yes, it produced a decision, and the decision should have been written down separately.

Three practices

Turn the bots off for the last five minutes. Not to hide anything. Because the presence of a recorder changes what people are willing to say, and the last five minutes need the version of the room that argues.

Write the decision before you write the actions. Actions are the easy part and they feel like progress. A list of actions with no decision behind it is a list of people being busy in different directions.

Name the reversal condition. What would have to be true for this to be reopened. It takes thirty seconds and it saves the same conversation in six weeks.

What this is really about

Here is the part that does not get said often enough about meetings and AI.

The meeting is not a container for conversation. It is an instrument for producing commitments that a group of people will act on without being chased.

When the record becomes the output, the instrument changes shape. People arrive prepared to contribute content rather than to decide. The meeting feels productive because it produced something, and the something is a document.

A decision is different. It has an owner, a cost, and a date after which it is too late. Those three things are uncomfortable, which is exactly why a good summary is such a tempting substitute.

The experiment

Take the last meeting you ran and try to write the four lines from memory.

What was decided. Who owns it. What would reverse it. When you check it.

If you cannot fill all four, the meeting produced a record. That is worth knowing, and it is worth fixing before the next one.

One page. Four lines. Five minutes. That is the whole change.

Where this connects

[Ownership Is a System](/ownership-system) describes the conditions that make a decision stick: clarity, authority, visibility. A decision minute is a small, practical version of all three.

[Pipeline vs Forecast](/pipeline-forecast) is the same problem in a different room. A forecast that looks complete and a meeting that looks decided fail in the same way, and they fail quietly.

If decisions in your business are recorded but not made, the [Revenue System Audit](/revenue-system-audit) is where we start.

Frequently asked questions

Should we stop using AI to take meeting notes?

No. Recording and summarizing are the parts AI does better than we do. The problem is not the record. It is treating the record as evidence that a decision happened.

What is a decision minute?

A one-page, four-line record written in the last five minutes of a meeting with the bots off: what was decided, who owns it, what would reverse it, and when it gets checked.

Does this work for remote meetings?

It works better remotely, because the summary is usually the only shared artifact. The decision minute gives the meeting a second output that the summary cannot replace. ## A five-minute reflection What did your last meeting decide, and who would notice if it had decided nothing? ---

Jerald Lee, Commercial Systems Advisor

Jerald Lee

Commercial Systems Advisor

Jerald Lee is a Hong Kong-based Commercial Systems Advisor, founder of Curiosity At Work and Operating Partner of The Growth Coach Hong Kong. He spent 13 years at Google APAC and helps founder-led companies build predictable revenue systems.

More essays from Curiosity At Work