Meeting notetakers work. That is worth saying first, because the rest of this is a criticism and the criticism is not that they fail at their job.
Fireflies, Granola, Otter, the one now built into your video calls: they join, they transcribe accurately, they summarize well, and they pull out a list of action items with names against them. Five years ago that was a research problem. Now it is a checkbox in a settings panel.
And the action items still do not get done. Not most of them, and not by most people. The failure has moved somewhere else, and it is worth being precise about where, because the obvious fix is the wrong one.
The list is in the wrong place
At the end of a call your notetaker produces a document. The document contains a summary, a transcript, and a short list of things that are now owed, including by you.
Then the next call starts.
The document is filed by meeting. Your work is organized by day. Those two things never meet again. To act on the list you would have to go back to the meeting it came from, and nobody goes back to a meeting. The notes are read by someone who missed the call and then, in almost every case, never opened again by anyone.
So the extraction was perfect and the outcome is the same as if nobody had taken notes at all. The commitment is written down in a place that is chronologically indexed, in a tool you open when you are in meetings, about work you do when you are not.
Why a shared summary is not a shared record
There is a second failure that is quieter and does more damage.
The notetaker's list belongs to the meeting, not to a person. It says that Ahmed will send the revised scope and that you will confirm the budget line. It is genuinely useful as a record of what was agreed, which is why people forward it around.
But nothing in the system knows the difference between a thing you owe and a thing you are owed. It is one list, presented flat. Both halves decay at the same rate, which is to say immediately, and neither half turns into pressure on anybody.
That distinction is not cosmetic. What you owe needs to reach your own week. What you are owed needs to reach a follow up, on a date, to a named person. They are two different mechanisms with two different failure modes, and a bulleted list under a heading called Action Items is neither of them.
The integration answer, and why it stops short
The usual response to all of this is an integration. Push the action items into Asana, or Notion, or a task manager of your choice.
This helps, and it is worth turning on. It also inherits two problems.
The first is that it fires on every meeting, so your list fills with the residue of calls that produced no real obligation. Most meetings produce zero genuine commitments and a notetaker that returns zero items looks broken, so the incentive runs toward generosity. After a fortnight of that you have a task list you have started skimming, which is the beginning of not having a task list.
The second is that arriving in a task manager is not the same as being enforced by one. A task manager holds what you put in it and shows it to you when you go looking. That is the right design for work you already decided to do. It is not a mechanism for work that got agreed to verbally, by you, in front of someone who is now expecting it, and who will remember.
What we do differently, and what we do not
Wasp reads meeting transcripts too. There is no cleverness to claim on the extraction itself, and pretending otherwise would be easy to check.
What happens to the result is different in three ways.
It splits the list by direction. Every item the model finds carries an owner. Something you committed to becomes a promise, tracked against you. Something someone else committed to becomes a waiting on item, tracked against them, with a follow up attached. A commitment between two other people in the room becomes nothing at all, which is the correct answer and is not what a meeting summary does.
It holds a higher bar for speaking than for writing. The threshold for accepting a commitment out of a transcript is stricter than the one for email, because speech is looser than writing. People think aloud in meetings. Wasp would rather miss a soft one than file somebody else's half thought as your obligation.
And it keeps promises out of your task list unless you say otherwise. A promise made to a person is a different object from a task you set yourself, and collapsing the two is how both stop meaning anything. Promises sit in their own place with the person's name on them. If you want one to become a scheduled piece of work, you promote it in one click. If you do not, it still gets chased, because it still exists.
The honest limit
Wasp only sees a meeting when a notetaker recorded it and Wasp is connected to that notetaker. If the call was on a platform whose built in notes never leave that platform, none of this applies and your commitments from that call are wherever you left them.
That is a real gap and it is a growing one, since every video product is now shipping its own summarizer that only talks to itself.
The general point survives it though, and it is the one worth taking even if you never use any of this. Getting the action items out of the meeting was the part that got solved. Getting them into the week where the work happens, in a form that distinguishes what you owe from what you are owed, and that keeps asking after the list has scrolled away, is still open. A better summary does not close it. Nothing about a document closes it.