Granola, Otter, and now Wispr have made AI transcription the assumed presence in business meetings. The consent question is still open.
Somewhere in the second half of a meeting, a person glances at the corner of their laptop and notices a small new presence in the room: an AI is transcribing everything being said, "mm-hmm" sounds and all. The meeting has a third listener now, and nobody asked it to sit down. That quiet recognition is the doorway into what actually changed.
Granola AI has been the Silicon Valley default for transcribed meetings for over a year. Otter AI has been around long enough that "did Otter join?" is a common pre-meeting check. This week, Wispr Flow, best known until now for voice dictation, joined the category with its own Notetaker, and its CEO framed the moment plainly: "This is the year of meeting recorders and notetakers," Tanay Kothari told Wired. The claim is hard to argue with, at least in Silicon Valley business meetings, where the social default has already tipped.
What tipped it is friction, not policy. Wispr's Notetaker connects to the user's calendar and listens along during any meeting, then generates a summary and surfaces an "ask anything" tool for follow-up questions against the transcript. Maximum meeting length was extended to what the company calls "a whopping six hours", up from a six-minute cap that users worked around by restarting the recorder five times to capture a half-hour. The Mac version launched this week; a Windows build is on the roadmap. TechCrunch's separate coverage confirmed the launch and an updated terms page on the same day.
Wired's reporter, Lauren Goode, writes that most tech investors and experts she has spoken with this year now assume business conversations will be recorded or transcribed by AI, sometimes without explicit consent. That is reported observation from a limited Silicon Valley sample, not a statistical claim about every office, but it tracks with what the products do by default: connect to a calendar, join a meeting, start transcribing, and save a searchable transcript. Wispr does ship an account-level Privacy Mode that prevents dictation data from being used for AI training. It is a real setting, not a full answer to the consent question.
The consent question runs underneath the product story. A meeting used to be a private space bounded by who was in the room and what they chose to remember. The default was off. The new default is on, attached to a calendar, and shared between the people on the invite and the vendor that wrote the transcription model. Nobody on the call asked for the third presence, and most of the time nobody announces it either. That is a different deal from a person taking notes with a pen, even if the words on the page look the same.
Three questions are worth asking before the next meeting: who is listening, what is being kept, and whether the host announced the recording. The first is the easiest to answer, because the answer is almost always an AI, and almost always named. The second is harder, and depends on the vendor's retention policy, the user's account settings, and whether training has been turned off. The third is the social one, and it is the lever: a host who names the recorder and checks the room changes the default for that meeting. Wispr shipped its Notetaker on Mac this week with a six-hour cap. A Windows build is on the roadmap. The recorder is already in the room. The announcement is not.