An AI meeting assistant that answers while the call is still going

At a glance

What it answers from

Ask the way you would ask a colleague who was at every meeting: “Hi Dot, what did we agree about the launch date last time?” Dot answers from what it can read, closest to the call first:

• What has been said in this call so far.
• What is shared in the meeting window, which Dot notes at most every ten seconds. On a Mac this needs the Screen Recording permission, and on Ubuntu an Xorg session rather than Wayland.
• The notes of your last five meetings, of any kind.
• The documents you added to the meeting, such as the agenda or a deck, and on a paid plan your organisation’s library.
• The web, only if your organisation has turned web search on, and then only the question is sent.

Where the answer appears

On your screen, in Dot. Anybody Dot can hear may say “Hi Dot”, and whoever asks, the answer is written for you, from what you have access to. Dot never speaks in the meeting.

A small Dot overlay floats over the call with the agenda item, the latest caption and a box to ask in. It is kept out of screen shares, so an answer meant for you is not shown by accident.

When the room should see an answer, switch Dot to Presenter and share its window in the meeting. Presented, Dot draws on earlier meetings only if they belong to the same recurring meeting, so nothing from your other calls reaches the room, and its window carries a line saying the meeting is being transcribed.

Before, during and after the call

• Before: link your Google or Microsoft calendar and Dot shows your next meeting a minute before it starts, with Join & listen in. Add the agenda, a deck or notes to a meeting, and Dot reads them first.
• During: answers when you ask, a board of what the room is deciding and who is taking what on, and suggestions only you see, such as a question somebody asked that Dot can answer, or a check on the time when an agenda item has one.
• After: the notes, with the outcome in one line and a name for a meeting that had none. Correct them, email them to colleagues, share the meeting or export the notes.

What Dot does not do

• Join the call. It is not in the participant list, and nobody has to admit it.
• Speak. Its answers are written, on your screen.
• Keep the audio. The call is transcribed as it happens, and the words are what is kept.
• Identify anybody by their voice. Names come from the meeting app’s own display of who is speaking, and a voiceprint is never stored.
• Tell the other people on the call. Dot shows them nothing, so that is up to you, and in many places the law requires it.