A named bot joined the call, and somebody in the room asked who invited it.

It happens on a client call more often than any vendor wants to admit: a notetaker bot logs in a few seconds after everyone else, sits on the participant list under its own name, and somebody — a client, a candidate, a new hire meeting the team for the first time — asks what it is and whether they should be worried. The meeting continues, but the moment has already landed slightly wrong, and it was entirely avoidable, because recording a meeting has never actually required a separate participant to do it.

Which tools do this, specifically

Otter, Fireflies, Zoom's own AI Companion and Microsoft Teams' Copilot recap all join a call as a named bot by default, which is how each of them gets access to the individual audio feeds it uses for its transcript. That is a genuine design choice with a real upside — the bot can be scheduled onto a calendar in advance and join a recurring call without a human remembering to start anything — and it works reliably for a team that never minds the bot being visible. It is worth naming plainly rather than implying every notetaker works this way: Granola specifically does not put a bot in the call either, reading the computer's own audio instead, the same underlying approach this page is about. The distinction that still separates it from what follows is where that audio goes afterward, not whether a bot showed up to begin with.

Reading what the Mac already hears, instead of joining anything

Coii AudioNotes records the microphone and the system audio the Mac is already playing, as two separate tracks — the same mechanism a screen recording uses, not a participant logging into the call. Nothing appears on anyone's screen, because nothing new is there to appear.

The transcript and the summary still get made

Skipping the bot does not mean skipping what the bot was there to produce. The transcript builds segment by segment while the meeting runs, and a language model bundled inside the app turns it into a summary once the call ends — the same outputs Otter or Fireflies would hand back, generated from the Mac's own recording rather than from inside the meeting.

Speakers still get told apart, without a bot's individual feeds

A bot in the call separates speakers because it has access to each participant's individual audio feed directly from the meeting platform. Reading the Mac's own audio instead means that trick is not available, so speaker separation works by voice print — matching each voice against the others already in the same recording. The result on the page looks the same, labelled lines like the transcript above; the mechanism producing it is different because there was no bot inside the call to lean on.

What still has to be said out loud

None of this changes a separate, real requirement: whether the people on the call need to be told they're being recorded. A bot's presence on the roster used to serve as a visible cue for that, however imperfect; without it, telling people plainly at the start of the call matters more, not less. The rules for that differ by jurisdiction and between one-party and all-party consent states — a question for your own counsel, not a claim this page is in a position to settle.

Why the bot got invited in the first place

It is worth understanding the actual reason bot-based tools default to joining the call, rather than treating it as a careless choice: a bot inside the meeting gets a clean, individual audio feed for every participant straight from the video platform, which is the easiest way to build a reliable transcript with speaker labels. Reading a Mac's own microphone and system audio instead means giving that convenience up in exchange for not needing a participant at all — a fair trade for someone who has been asked about the bot once too often, and a real reason a team that has never minded the bot might not bother switching.

What it deliberately doesn't do

There is no calendar integration that logs a bot into a recurring meeting automatically — recording starts when you start it on the Mac that is already in the call, which is a small manual step in exchange for nothing showing up uninvited. There is no shared workspace for a colleague to read the transcript from their own login; the recording and its transcript stay on the Mac that made them, across up to three machines on one licence. And it is built for macOS specifically, with no Windows or Android version, so a team meeting across mixed operating systems will not get a consistent bot-free experience from this one tool alone — worth knowing before assuming it solves the problem for everyone on a call, not just the person running it.

The trust question underneath the roster

There's a version of this that goes beyond a single awkward moment: a client, once they've noticed a bot on one call, sometimes starts wondering what happens to the recording afterward — who has access to it, how long it's kept, whether it's used for anything beyond the meeting itself. A vendor's own privacy page can answer that eventually, in a document nobody in the room has actually read, but the question usually surfaces mid-meeting instead, at the exact moment nobody is in a position to pull that page up and read it aloud to a client waiting for an answer. Not having a separate party on the call in the first place sidesteps the question entirely, rather than needing a good answer ready for it.

What actually changes on an ordinary call

Nothing about the meeting itself changes — the same platform, the same join link, the same conversation. What changes is entirely on the host's own machine: instead of waiting for a bot to log in and confirming it joined correctly before starting to talk, recording begins the moment the call does, silently, with nothing on screen for anyone else to notice or ask about. The first few calls after switching away from a bot-based tool are usually the ones where the habit of glancing at the participant list to check a bot joined has to be unlearned — a small adjustment, and the last thing that actually changes about how the meeting itself runs.

What it costs, what it runs on

Coii AudioNotes is $19, paid once, for three of your own Macs, with a 30-day trial needing no card and no account. It runs on macOS 13 Ventura or later, Apple Silicon or Intel. Six people moving off Otter's or Fireflies' monthly subscriptions land at $114 total rather than roughly a thousand dollars a year, on top of no longer needing to explain a bot to a client every few calls. Full comparisons against each bot-based tool named above: Otter, Fireflies, Zoom AI Companion and Teams Copilot recap. Against the other bot-free option on this list, Granola shares the no-bot design but remains a cloud subscription rather than a purchase that runs entirely on the Mac.

What "no bot" changes beyond that one awkward moment

The client asking who the bot is turns out to be the most visible symptom of a wider pattern, not the whole of it. A bot has to be given calendar access or an invite to every meeting it's meant to join, which means a one-off call — a candidate interview, a first conversation with a prospective client, a call booked five minutes before it starts — often gets missed entirely if nobody remembered to add the bot in time. Reading the Mac's own audio sidesteps that failure mode along with the awkward-moment one: recording starts because a person pressed record, not because a calendar event was configured correctly in advance, which means the unscheduled call gets captured just as reliably as the recurring one.

Recording is a decision, not a default

There's a trade worth naming honestly in the other direction, too. A bot that auto-joins every meeting on a calendar captures everything without anyone having to remember to do it, which is a genuine convenience for a team with dozens of calls a week and no consistent habit of hitting record. Moving to a tool that has to be started manually shifts that responsibility onto the person in the meeting — a real cost for a team that valued the bot's automatic behaviour specifically, and worth weighing honestly against the awkward-moment problem rather than assuming the trade is free in either direction.

The moment it's easiest to actually notice the difference

The clearest way to feel the difference is a call that was already going to be a little tense — a renewal negotiation, a candidate who's clearly nervous, a first conversation after a project went sideways. Those are exactly the calls where a small, unexplained thing on screen gets noticed more than usual, because everyone in the room is already paying closer attention than normal. Running the same call with nothing but the usual participants on the roster removes one variable from a conversation that already has enough of them, without changing anything about what actually gets recorded or written up afterward.

Who this genuinely isn't built for

A team that mostly meets internally, never in front of a client or a candidate who might notice or question a name on the participant roster, has genuinely less reason to care about the bot itself — the awkward-moment problem this page opens with simply doesn't come up for them as often, and a bot's automatic scheduling may be worth more than avoiding a scenario that rarely happens. This is written for the specific situation of an external call where a bot's presence has actually landed wrong at least once, not as a general claim that every meeting, internal or external, benefits equally from skipping it — an internal stand-up between two colleagues who already know each other well rarely has anyone in the room who would even notice the difference either way.

The test that actually settles it

Run one real client call with the bot-free option instead of whichever notetaker currently joins by name, and watch whether anyone even notices a difference in the meeting itself — that absence of a reaction is the entire point. A round-up of the other tools built around the same complaint is here, and the wider case for what "AI notetaker" means once the bot is no longer part of the definition is covered here.