Wispr Flow Wants the Minutes Without the Clerk

Wispr Flow has released Notetaker, a Mac tool that connects to a user’s calendar, listens during workplace meetings, displays a live transcript and produces an AI summary afterward. Users can question that record through a chatbot-style interface. Wispr says transcription occurs on the device, and an optional Privacy Mode prevents dictation data from being used for model training.

The product enters a busy market that already includes Granola and Otter. Its sales proposition is familiar: stop assigning a person to take notes and let the meeting document itself. In an initial test reported by Wired, the transcript and summary showed no obvious errors, and a quoted remark matched the audio when checked. That check is the important part.

A transcript does not certify itself. Someone still has to confirm who spoke, distinguish a proposal from a decision, recover a name the software mangled and compare a disputed sentence with the recording. The clerk has not disappeared. The work has been broken into smaller tasks and returned to everyone, usually without a title or allotted time.

Consent follows the same pattern. Wispr can transcribe directly on a laptop rather than entering the call as a visible bot, which makes the tool less intrusive on screen and potentially less obvious to other participants. A workplace therefore needs a rule stronger than assumption: who announces recording, whose agreement is required and what happens when a client, applicant or employee declines.

Storage creates the next shift. A six-hour recording can contain personnel discussions, customer information, legal advice and remarks never intended to become a searchable institutional file. Privacy Mode addresses model training, but employers still need retention periods, access controls and deletion procedures. “The vendor offers a setting” is not a records policy.

Before the bot joins, an employer should name the person responsible for consent, verification, correction, access and deletion. Workers also need a clear way to challenge a summary before it becomes the accepted account of a decision. Automation can draft the minutes. It cannot decide who answers when the minutes are wrong.