The transcript is evidence, the summary is the artefact
The useful output of a recorded meeting is a page you can read in a minute. The transcript sits underneath it, and its job is to be there when someone disagrees about what was said — not to be read end to end.
People who get value out of recording read the summary and search the transcript. People who do not are trying to read the transcript, once, and then never recording again.
Extract three things and ignore the rest
Decisions: what was settled, and by whom. These are what you will be asked about in a month.
Commitments: who said they would do what, and when. This is the highest-value extraction, because a commitment made out loud with a date is the thing most likely to be dropped.
Open questions: what was raised and not resolved. These are the agenda for the next meeting, already written.
Send the summary within the hour
A summary sent while people still remember the meeting gets corrected. One sent two days later gets ignored, and an uncorrected summary is worse than none because it becomes the record.
Send it to everyone who was in the room, including the people who said nothing. They are frequently the ones who notice the mistake.
Search is the feature that pays for itself
The value of recording compounds. One transcript is a nice-to-have; two hundred of them are a searchable record of every commitment anyone has made to you.
The question that makes it worth it is not "what happened in that meeting" but "when did we agree that, and who was in the room" — which is unanswerable from memory and trivial from a search box.