The AI Meeting Singularity

Meetings are an inherently social affair. A group of people comes together to share information, discussion, bond, build and maintain trust, share feedback and ideas, and, ideally, be candid and productive with each other. In a good meeting, decisions are being made and everyone walks away knowing exactly what’s been decided. Should anyone forget, there’s an artifact that everyone can go back to and remind themselves and each other. That’s your set of meeting notes.

They are also the most tedious part of a meeting. Taking good notes is a skill, one you can learn, but a skill nonetheless. More often than not, meeting notes are a stream of consciousness of everything that’s been said. Summarizing them into something meaningful is an extra step. Which is exactly why AI note taking are so tempting.

The promise is simple. You can analyze how much of a meeting someone’s taken up. Decisions bubble up through the AI summary almost effortlessly and without anyone truly having to say: that’s what we’ll do. The transcript is the single source of truth and can always be queried. Information, decisions, and themes can be tracked and traced across several meetings, as all of them have been transcribed. And I will grant: this is truly one of the useful things of having AI-enabled tooling in a business, the ability to query a disjointed set of data to retrace history.

The reality is though, and AI changes little about this fact, that meetings are, by and large and in the view of the majority of workers, a waste of time. In an Atlassian survey ^1, almost 80% of people surveyed said they’d get more work done if they were in fewer meetings. The same survey suggests that the majority of meetings is ineffective. In addition, the shift towards more remote or hybrid work setups has only contributed to an increase of the number of meetings being held ^2.

It seems only natural that we turn to tooling to help us making them better, or at least give us the illusion that they are better. After all, with an increasing number of meetings, there’s less time to prepare and post-process, like preparing and distributing meeting notes.

Plus, throwing AI note-taking tools in the mix has the added benefit of pleasing the boss who has mandated increasing use of AI to increase productivity.

AI note-taking during meetings isn’t the end-all solution, though. More meetings, more notes, and more ways to query meetings don’t necessarily translate to more or better decisions being made. It seems more likely that decisions will be left to the AI tools to make for us.

Having presumed access to the entire history of meetings leaves us with an illusion. It seems convenient to be able to go back in time and extract meaning from a transcript. But meaning isn’t just in the words being said.

Information is also in who’s speaking and who isn’t. It’s in eye-rolls, in silent nods, shaking heads, in somebody unmuting to say something only to mute themselves again. Information is in whether the boss is present in the meeting, independent of whether they’re saying anything. Their mere presence can change the dynamic of a meeting, an effect we call “Schrödinger’s Meeting” in our book, “The Intentional Organization,” named after the infamous though unnecessarily gory thought experiment by Erwin Schrödinger.

Information is also sometimes spread out over many spoken sentences, intermingled with yeah’s, no’s, and ”hum’s. Meaning may only emerge when several people have spoken and one ore more people present finally put the strings together. Sometimes it’s those emergent threads that come up only in meetings or in their wake that truly shift perspectives and lead to good decisions. Sometimes good ideas simply aren’t said out loud, for whatever reason. It’s a thread to be picked up on later, maybe just for yourself, or for the benefit of the team. But without anyone taking notes, those things are easier left unsaid.

We are, after all, people driven by our intuitions and associative brains. AI doesn’t have access to those. Going back to transcripts and trying to pick up that kind of a thread in a free-flowing discussion can be next to impossible. That’s why I always, even when I use a note-taking tool, keep a paper notebook next to me. Key ideas, especially those that may not immediately concern the other parties, go in there. Threads I want to pick up on later are written down as well.

AI tools do one thing very well: they give us the illusion of access to an abundance of knowledge. In the case of meetings, they allowing us to travel back in time and look for patterns, allowing us to piece together things we might’ve missed, or maybe because we couldn’t attend the meeting.

That is indeed a neat feature, but having access to all that kind of knowledge gives us the illusion of knowing. Just because we can know things doesn’t mean we do.

Having access to knowledge isn’t the same thing as knowing.

Just because we could go back and analyze a meeting in hindsight doesn’t mean we know what actually went down. It just gives us an illusion of control, that we can be present in many more meetings without actually attending them, because our schedule is already so full with meetings. The snake bites its own tail.

One last thing to consider is that people may actually be holding back in meetings. When things are transcribed, it can be all too easy for anyone to go back, pick and choose words and twist them, or use them with ill intent. People might not share what they think because they fear it will be used against them at some point. When everything is recorded, something you said five years ago might resurface at a point where you stand at odds with a new boss or when the decision is up for your next promotion.

All this said, there is indeed value in having these transcriptions available and accessible. But the question is why you need both in the first place. When there are proper notes being taken, decisions and follow-ups documented, when there are clear roles in a meeting (like facilitator, moderator, note-taker, timekeeper), you may not need these transcripts in the first place. What AI surfaces from an increasing amount of data will always be an easily digestible summary. But when information is averaged into only a few paragraphs, meaning gets lost, which arguably defeats the purpose of having these tools in the first place.

AI note-taking tools can certainly help in keeping a historic record of everything that’s been said. But they might just remove the impetus for asking the question that, according to the research at the beginning of this article, isn’t asked often enough: do we even need to have this meeting in the first place?