I counted it once and wished I hadn't. Six recurring meetings a week. Four people in most of them. That's roughly twenty-four human-hours a week spent listening to each other read updates we could have typed.
So I ran an experiment. For one month, I cancelled every status meeting and handed the job to an AI instead. No standups. No "quick syncs." No calendar invites with eleven people and one decision.
I expected chaos. What I got was quieter, and a little uncomfortable, and mostly better.
You can replace status and update meetings with AI almost entirely, but you cannot replace decision meetings or conflict meetings. The trick is sorting which is which. Async written updates, summarized and routed by an AI assistant, killed about 70% of my meeting time without losing information. The other 30% got better because the people who showed up actually needed to be there.
Here's how the month went.
I didn't build anything fancy. People worried I'd spent a weekend wiring up some monster of an automation. I didn't.
Each morning, everyone dropped a three-line update into a shared channel: what they finished, what they're doing today, what's blocking them. An AI assistant read all of them, grouped them by project, flagged anything that looked like a blocker, and posted one clean digest by 9:30.
That was the whole system. Updates in, one summary out.
The first surprise was how much shorter people's updates got when they knew a machine would read them and a human would skim them. No performance. No "just to give a bit of context." Just the facts.
I went in skeptical. By week two I was a little annoyed at how well some of this worked.
The AI was genuinely strong at:
That last one mattered more than I expected. Most meetings exist partly because nobody wrote anything down last time.
Now the honest part. There were things the AI could not do, and pretending otherwise would make this a worse article.
It could not read the room, because there was no room. When two people quietly disagreed, the AI summarized both positions neatly and moved on. It had no idea one of them was frustrated. A blocker that's really a relationship problem looks, in text, like a scheduling problem.
It also couldn't make a call. When a real decision needed an owner, the digest would say "needs a decision" and then sit there. The AI is excellent at surfacing the question and useless at answering the kind of question that involves taste, risk, or someone's feelings.
The meetings worth keeping are the ones where someone has to decide something, and someone might be wrong.
So I kept exactly those. One weekly decision meeting, thirty minutes, only the people who owned a live decision. Everything else went async.
I'm labeling these as my real counts from the month, not a study.
| Meeting type | Before (hrs/week) | After (hrs/week) |
|---|---|---|
| Daily standup | 5 | 0 |
| Project status syncs | 6 | 1 |
| "Quick" syncs | 4 | 0.5 |
| Decision meetings | 2 | 2 |
| Total | 17 | 3.5 |
That's about thirteen hours a week back across the team. Not because we did less. Because we stopped narrating the work out loud.
Async plus AI is not free. It moves the cost somewhere else.
Writing a good update is harder than mumbling one. A few people on the team genuinely missed the meetings, and not for lazy reasons. The standup was where they felt like a team. Take it away and some people feel like freelancers who happen to share a logo.
So we added one thing back on purpose: a Friday call with no agenda and no updates. Just people. The AI handled the work; the humans handled being humans. That balance is the actual finding here.
If I'd only automated and never added that back, I think morale would have quietly leaked out over a few months. Automation that ignores the social half of work tends to look efficient right up until people start leaving.
If you want to copy the experiment, here's the order I'd do it in:
This is the same instinct behind a lot of good automation: let the AI handle the repetitive, written, searchable stuff, and give the humans back the hours for judgment. It's the throughline in the honest truth about AI productivity tools, and it tracks with what McKinsey's research on AI adoption keeps finding: the gains show up where automation removes routine work, not where it tries to replace human decisions. I leaned on the same split when I later wrote about cutting my meeting time roughly in half.
The experiment ended, but I didn't go back to the old calendar. A few beliefs got rewired permanently, and they're worth more than the hours I saved.
The first is that most meetings are a failure of writing. We meet because nobody wrote the thing down, or nobody read what was written, or the writing was so bad that talking felt easier. Fix the writing and a startling number of meetings simply evaporate. An AI assistant that turns sloppy updates into a clean digest is, weirdly, a writing tool first and a meeting tool second.
The second is that presence is a budget. Every meeting spends a little of everyone's attention, and attention doesn't refund. Before the experiment I treated meetings as free because they were "just talking." They were never free. They were the most expensive thing on the calendar, paid in the one resource you can't buy back. Once I started seeing each meeting as a withdrawal, I got much stingier about scheduling them.
The third surprised me most: async made people more honest. In a live meeting, social pressure smooths everything over. Nobody wants to be the person who says the project is behind. But a written update, read by a calm AI digest, has no audience to perform for. People reported problems earlier and more plainly. We caught two slipping projects weeks sooner than we would have in a room full of nodding.
If this resonates, it's worth trying the "update, decision, or human" sort on your own calendar for a single week and seeing what survives.
I also learned the limits of my own optimism. For about a week I genuinely believed I could automate nearly everything and run a frictionless, meeting-free team. That was naive. The human meeting I added back wasn't a concession — it was load-bearing. The lesson isn't "automate meetings." It's "automate the information, and protect the connection." Those are two different jobs, and confusing them is how teams quietly fall apart while looking efficient on paper.
Q: Did people just stop reading the AI digest? Some did, at first. The fix was making it short and putting the blockers at the very top. A digest people actually read is one they can finish in under a minute.
Q: What about brainstorming? Doesn't that need a meeting? Brainstorming is a "human" meeting, not an "update" one. I kept those. The point was never to delete meetings. It was to stop using meetings for things writing does better.
Q: Isn't this just Slack with extra steps? The AI summary is the difference. Raw channels create more noise, not less. The assistant turning twenty messages into one digest is the whole value. Without it, you've just added a chat tab.
Q: Did anything important get missed? Once. A blocker got buried because someone phrased it as a side comment. We fixed it by telling people to label blockers explicitly so the AI could catch them every time.
Most meetings are updates wearing the costume of collaboration. An AI can run the updates better than a room full of people can, but it can't decide, and it can't make anyone feel like they belong. Automate the first part. Defend the second.
Meetings should be for the conversations that can't be summarized.
If you ran this experiment on your own calendar, how many of your recurring meetings would survive the "update, decision, or human" test? Worth finding out before you book next week.
One person, output that looks like five. It isn't about working more hours — it's about a kind of leverage teams rarely have.
One idea a week to a published issue in under an hour. The boring system behind a newsletter I never dread sending.
Behind a lot of lean, profitable companies is the same small stack of AI tools. Here's what's actually running the show.
Comments
Sign in to join the conversation
No comments yet. Be the first to share your thoughts!