Shameless plug as excitingly I'm working on a tool https://designpro.ai which converts input from various sources into insights and tasks. I've used it to generate insights from my call transcripts, so I do know that it does work.
I moved to google workspace precisely because of the ease in which you can record and retrieve recordings of meetings. (IE; it's embedded into the calendar event).
If a meeting is not important enough to be recorded then it probably wasn't a meeting worth having.
My personal position is that I put more hours into the day than other people, but it's no synchronous, I might be in a more important meeting when there's an important meeting; a lot of nuance gets taken out if you just read meeting notes (which people seldom do also) - but even there it's so helpful in revising meetings notes quickly.
I really push quite hard for having meetings recorded.
The necessity of re-watching to identify inconsistencies in a meeting is a problem with the medium. An inconsistency in thinking early in a meeting can invalidate the whole meeting but because a meeting has to move at pace it's much more likely for an inconsistency to slip in. Team meetings also bias the conversation towards whoever is most confident in speaking up, this is especially disadvantageous towards people new to a team or those where the dominant language is not their own.
I'm not against meetings outright as some people much prefer communicating using their voice instead of in writing, and a good team is one that is able to accommodate each member's preferred style of working... but if a team meeting is taking place that is both important and also stretches the capabilities of participants, it should not be a meeting. A meeting should only happen when every participant has decided that it's the best medium for that specific conversation.
A meeting is like sitting in front of a hose: details will be missed.
Asynchronous meeting participation is fantastic for all the reasons stated above. It makes the information in the meeting discoverable. Meetings can now be captioned by NLP, which makes the content discoverable to LLMs. We currently train an internal chatbot on our Confluence content and are extending that to train it on the content of recorded meetings.
Then, the next day, I can listen to the call, pause, increase speed 2x,3x, take notes for myself, write minutes, write summaries for internal use (casual slack message), notice inconsistencies in our thinking, or run an AI note taker all on my own schedule.
And let's not forget when someone is sick, need to pick up a mother in law from the airport, have a parallel meeting, or joined a week after an important alignment.
If the meeting wasn't that important? No problem, just ignore it. Set up a sane auto delete rule, and that's it.
In practice, I don't think people spend time snooping around where they shouldn't, so that isn't really an issue.