Every Thursday at 2 PM, twelve people sit in a meeting called “Cross-Functional Sync.” It runs for sixty minutes. I attended every one of them for the first eight months I was at the company. Then I built an AI workflow to summarize them. Then I stopped attending.
It has been six weeks. Nobody has noticed.
How the Workflow Started
I did not set out to skip the meeting. I set out to stop taking notes during it, because the notes I took were uniformly useless: half-sentences I could not parse two days later, action items I had no context for, names of people whose roles I had to look up.
I asked our IT-approved transcription tool to record the meeting and pipe the transcript into Claude with a single prompt: “You are reading the transcript of a one-hour cross-functional sync. Identify any decision that was made, any commitment a named person made, and any topic that was deferred to a future meeting. Ignore everything else.”
The output, the first time, was about 200 words. The meeting itself was 60 minutes of audio.
What the 200 Words Showed Me
The transcript-to-summary ratio told me something I had been suspecting for months. The actionable content of the meeting was about three percent of the meeting’s runtime. The other 97 percent was context-setting, status updates that were already in the weekly email, and tangents that did not produce decisions.
After four weeks of summaries, I noticed that I was reading the summary on Friday morning, taking the one or two action items relevant to me, and forgetting the rest. The meeting itself had become an artifact I was ignoring while I waited for the summary.
So one Thursday, I did not attend. I read the summary on Friday. My week was identical.
I tried it again. Same result. I have now missed six in a row.
What This Says About Meetings
The most uncomfortable part of this experiment is not that I stopped attending. It is that nobody noticed.
That tells me one of two things. Either my attendance never mattered, which is plausible, but I would expect at least one person to ask where I had been. Or the meeting is operating below the threshold where individual attendance is tracked, which suggests the meeting itself is operating below the threshold where it produces value.
I think it is the second one. I do not think the meeting should exist.
But the meeting has existed for years, and a number of people would have to agree to end it, and the cost of attending is borne by twelve people while the benefit of ending it would be borne by no one in particular. So it will continue.
The Workflow I Actually Use Now
I do not summarize the cross-functional sync anymore, because I do not have a transcript. I am not in the meeting. Instead, I added a single prompt to my Friday morning routine: “Open the meeting Notion page. Read the agenda and any notes someone else took. Summarize in three bullets what I would have said if I had been there, and three bullets of what I needed to learn from being there.”
The first list is what I add to the Notion page as a comment. The second list is what I follow up on directly with the relevant person, in writing.
That is the entire system. It takes about twelve minutes on Friday. The meeting takes sixty minutes on Thursday.
The Honest Caveat
I am not recommending this. The reason this works is that I have been at the company long enough to know which meetings actually require my voice and which do not. A new hire could not make this call, and would correctly assume that all twelve attendees being present means all twelve attendees are needed.
The lesson is not “skip your meetings.” The lesson is that an AI summary of a meeting is also a measurement of the meeting’s information density. Once you have seen a few of them, you will not be able to unsee what you learn.
Some of your meetings will not survive that measurement.
Field Notes
- I shared the prompt with two colleagues. One started using it. The other found it depressing and asked me to stop talking about it.
- The IT-approved transcription tool is the same one I wrote about in The Only AI Tool I Have Not Uninstalled. The fact that it runs in the background is what made this whole experiment possible.
- I have not extended this experiment to meetings I run. I should. I am not sure I want to.