Should an AI Meeting Recorder Be Visible in the Room?
Granola’s new Apple Watch app can start transcribing an in-person conversation with one tap. The watch turns bright green and plays a sound, so the wearer knows it is working. That solves the wrong person’s uncertainty. A green wrist tells the recorder. It does not tell the person being recorded. Granola’s own consent guidance is clear for online meetings: disclose beforehand, ask verbally, and use a chat notice or video watermark. A walking one-on-one or candid coffee has no meeting chat and no shared screen. It needs an equally plain signal in the room—not a tiny status only the wearer can read. I’d want the watch to show a full-screen “AI notes on” card long enough to turn outward, then prompt for a spoken yes before the substantive conversation starts. If someone says no, stopping should be one tap and manual notes should remain easy. The watch removes the awkwardness of pulling out a phone. Good. It should not remove the moment when the other person gets a choice. What would count as clear notice in a walking meeting?
Comments
The hard test is whether a junior employee can tell their manager, “not for this one,” without having to explain why. Recording should stay off until everyone says yes, pause when someone new joins, and be easy to stop for one sensitive part of the conversation. If saying no changes who gets invited or whose version of the meeting counts, the green screen is decoration.
Granola’s consent page supports that boundary: it recommends asking before transcribing and puts the responsibility on the person using the app. The Watch launch only says the green screen and sound tell the wearer it is working; it describes no participant-facing notice or consent step. That leaves the junior employee depending on the manager to remember to ask. A safer default would keep transcription paused until consent is acknowledged, then let anyone stop it without having to justify why this part should stay out of the transcript.