What does 'AI meeting notes without a bot' mean?

Most AI notetakers (Otter, Fathom, Fireflies, tl;dv) work by sending a bot to join your video call. The bot appears in the participant list and records the meeting from inside the call. (Granola is the partial exception — it taps your Mac's audio rather than sending a bot, but it's still tied to the laptop and the scheduled call.) 'Without a bot' means the audio is captured locally — through a wearable pendant or your phone's mic — so nothing joins the call, nothing appears in the participant list, and the meeting runs exactly as it would have if no tool were there.

How does Fluent take meeting notes without joining the call?

Fluent captures the audio in the room rather than on the call. It listens through a paired Limitless or Omi pendant, or through your phone's microphone (phone capture is rolling out now). The recording lives in your own Fluent account, and Fluent produces a summary, action items by speaker, and a speaker-attributed transcript after the meeting ends. No bot is ever added to the call.

Which AI meeting-notes tools join the call as a bot?

Otter, Fathom, Fireflies, and tl;dv all rely on a bot that joins your Zoom, Google Meet, or Teams call as a participant to record it. Granola is the exception — it captures your Mac's audio locally instead of sending a bot, but it's still bound to the laptop. Either approach works for scheduled video calls but fails for in-person meetings, walking meetings, dinners, and hallway conversations.

Can Fluent take notes for in-person meetings?

Yes — that's the case bots can't cover. Because Fluent captures audio from the room (pendant or phone mic) rather than from a video call, it works for in-person 1:1s, walking meetings, lunches with a client, investor pitches in their office, and conference hallway conversations. Most of the conversations that actually matter never happen on a call a bot could join.

Where does the recording live, and is it private?

The audio lives in your own Fluent account, encrypted at rest, sharable only by you, and never used to train models. A bot-based tool routes your meeting through a third-party server that wasn't part of the room and whose vendor may train on the audio. Per-user encryption and on-device processing are on Fluent's roadmap and aren't claimed until they ship.

AI Meeting Notes Without a Bot

Q: Does Fluent join Zoom, Google Meet, or Teams calls?

No. Fluent never joins a video call as a bot or a participant, and it doesn't record the call stream. For a remote video meeting, Fluent captures the audio in your room — your side of the conversation and anything audible on your speakers. It will never appear in anyone's participant list.

Short answer: Most mainstream AI notetakers — Otter, Fathom, Fireflies, tl;dv — record your meeting by sending a bot to join the call. (Granola skips the bot and taps your Mac's audio, but it's still chained to the laptop.) Fluent captures the audio in the room instead, through a pendant or your phone's mic, so nothing joins the call, nothing lands in the participant list, and the meeting happens exactly as it would have without any tool present. That's the whole difference, and for in-person and walking meetings it's the difference between coverage and a blank page.

You've seen the bot. A few minutes into the call, a new participant slides into the list: Otter.ai Notetaker. Fathom is recording. A small gray square with a vendor's logo, sitting in the meeting like a court stenographer nobody invited. Everyone clocks it. The conversation tightens by a degree. The candid version of the meeting — the one where someone says the real reason the deal stalled — moves to the hallway afterward, where the bot can't follow.

That bot is not an implementation detail. It is the product. And it's worth asking why a tool that promises to fade into the background announces itself in the participant list of every meeting it touches.

How every other AI notetaker works

The category settled on one mechanism early, and almost everyone copied it. To get the audio, the tool joins the call.

Otter sends Otter Notetaker into your Zoom, Meet, or Teams call.
Granola leans on your Mac's audio but is built around the scheduled video meeting.
Fathom, Fireflies, and tl;dv each dispatch a recording bot that appears as a participant.

It's a reasonable engineering choice. The call platform exposes the audio stream; the bot taps it. But it bakes in three constraints that no amount of better summarization fixes:

It only works on a call a bot can join. No Zoom link, no notes. The meeting has to be a scheduled, software-mediated video call.
It's visible. The recorder is in the room, named, logged. People talk differently when they can see they're being recorded by a third party.
It routes your meeting through someone else's server. The audio leaves the room and lands in a vendor's cloud that was never a participant — and many of those vendors reserve the right to train on it.

For a calendar full of remote video calls, that's a tolerable trade. For the meetings where the important things get said, it's a structural miss.

What "without a bot" actually means

Fluent captures the conversation where it happens: in the room. It listens through a paired Limitless or Omi pendant, or through your phone's microphone — phone capture is rolling out now, so you no longer need a pendant to start. The audio is recorded locally to your own account. Nothing joins the call. Nothing appears in any participant list. When the meeting ends, the notes are waiting.

The mechanism sounds almost too plain to be a differentiator, until you notice what it unlocks. A tool that captures the room instead of the call isn't limited to calls. It covers the entire surface of your working day that the bot-joiners structurally cannot reach.

The meetings a bot can't follow

This is the part the funded incumbents can't copy without rebuilding their core mechanism. A bot follows a calendar invite. It cannot follow you into:

An in-person 1:1 where the real feedback gets delivered.
A walking meeting with a coaching client or a report.
Lunch with a customer where the renewal actually gets decided.
An investor pitch in their office, where you'd never ask to add a recording bot.
The conference hallway, where the most useful sentence of the week gets said and forgotten by dinner.

Most of the conversations that move your work forward were never going to happen on a Zoom link. Coaches live in walking meetings. Founders close in dinners and hallways. Managers do their best work across a table, not across a screen-share. A meeting-notes tool that only works on scheduled video calls is, for those people, a tool that works on the least consequential third of their week.

If you're weighing Fluent against a specific incumbent, the comparisons get specific: Fluent vs Otter comes down to the bot in the participant list; Fluent vs Granola comes down to the desk you're tied to. Either way, the conversation has to come to the tool — Fluent goes to the conversation.

The privacy difference is the same difference

The bot question and the privacy question turn out to be one question.

When a bot joins your call, your conversation lives on a third-party server that wasn't part of the room. The other people in the meeting may not have agreed to that vendor's terms. The vendor may train on your audio. The recording exists in a system you don't fully control, attached to a company that may be acquired, may change its policy, or may simply be breached.

When Fluent captures through your pendant or phone, the audio lives in your own Fluent account — encrypted at rest, sharable only by you, never used to train models. Per-user encryption and on-device processing are on the roadmap; we don't claim them until they ship. But the baseline is already different in kind: the recording never had to leave the room to reach a server that wasn't invited.

What you get instead of a transcript

Capturing the room is the unlock. The notes are the payoff. Fluent doesn't hand you a wall of transcript and call it a summary. After the meeting:

A two-paragraph summary, written the way a good chief of staff would write it — what the meeting was about, what got decided, what didn't.
Action items by speaker — not "follow up on Q3" floating in the void, but "Karim said he'd send the revised forecast by Friday." Owed by whom, to whom.
A speaker-attributed transcript — diarized, timestamped, searchable. Skip the parts you remember; find the exact line you misquoted.
Patterns across meetings — how much you talked, where the room went quiet, what got raised three meetings running before anyone moved on it.

And because the conversation now lives in a system you own, it can do things a transcript can't. Fluent's calendar agent turns the commitments in a meeting into follow-through. Its MCP server lets you point Claude or another AI at your own conversations and ask questions across all of them. The full picture is on the meeting notes page.

The one thing Fluent won't do

Honesty about the boundary, because it's the obvious objection: Fluent will not join your Zoom call and record the remote stream. That's the bot's job, and the bot is the thing we're deliberately not building. For a remote video meeting, Fluent captures the audio in your room — your side, plus whatever's audible on your speakers. If your entire job is recording the verbatim far-side audio of back-to-back remote calls, a bot-joiner is the tool built for that, and you should use one.

For everyone whose real conversations happen across a table, on a walk, over lunch, or in the five minutes after the call ends — the meeting notes were never going to come from a bot. They were going to come from the room.

The meeting happens. Nothing joins it. The notes are waiting when it's done.