The meeting bot problem

You've seen the bot

You're in a Zoom call with a client. The meeting starts. Then the notification appears: "Otter.ai's Notetaker has joined the meeting."

Everyone sees it. The client sees it. Your CEO sees it. The candidate you're interviewing sees it. A literal robot has entered the room, and now everyone knows the conversation is being recorded and transcribed by a third-party service.

If you use Otter.ai, Fireflies, tl;dv, or similar services, this is your life. Every meeting you record gets a bot participant. It shows up in the attendee list. It sometimes introduces itself in chat. On some platforms, participants get a notification that a "recording bot" has joined.

This isn't a bug. It's the core architecture of how these services work. And it creates problems that go far beyond awkwardness.

Why the bots exist

To understand the bot problem, you need to understand how cloud transcription services capture audio. They can't access your microphone directly — that would require installing software on every participant's device. Instead, they join the meeting as a participant, just like a human would, and record the audio stream from the meeting platform.

The bot is essentially a virtual attendee that sits in the call, captures the mixed audio feed, and sends it to the service's cloud servers for transcription. This is the only way to record a meeting without requiring every participant to install an app.

It's clever engineering. It's also terrible UX.

The social cost

Meetings are social interactions. They have norms, expectations, and power dynamics. Introducing a recording bot changes all of them.

The permission problem. When "Otter.ai's Notetaker" joins a client call, someone inevitably asks: "What's that?" Now you're spending the first two minutes of your meeting explaining that you use an AI transcription service, that it's recording the conversation, and asking if everyone is okay with that. Sometimes they're not. Sometimes the meeting starts with friction that didn't need to exist.

The formality shift. People behave differently when they know they're being recorded by a bot. They're more guarded. They hedge more. The casual, honest back-and-forth that makes meetings productive gets replaced by performative professionalism. The bot doesn't just record the meeting — it changes the meeting.

The perception problem. To external participants, a recording bot signals something about your organization. Maybe it signals efficiency. More often, it signals surveillance. Candidates in interviews see it and wonder what kind of culture they're joining. Clients see it and wonder where their confidential strategy discussions are being stored.

The enterprise problem

Many companies have explicit policies against meeting bots. Financial services firms, law firms, healthcare organizations, and government agencies routinely ban third-party bots from joining calls. Their reasoning is straightforward: the bot captures audio and sends it to external servers, which creates compliance and data sovereignty risks.

If you're on a call with a Fortune 500 client and an Otter bot joins, you might get a polite request to remove it. Or you might get a less polite email from their legal team afterward. Either way, the bot just created a problem that didn't need to exist.

Some meeting platforms are starting to push back too. Zoom has added controls that let hosts automatically block bots from joining. Microsoft Teams has similar restrictions in enterprise environments. The platforms recognize that bots are a friction point, and they're giving admins tools to eliminate them.

The infrastructure problem

Behind every meeting bot is a cloud server. The bot joins your call, captures the audio stream, and uploads it to the transcription service's infrastructure for processing. This creates a chain of data exposure that you can't opt out of:

Your audio travels from the meeting platform → to the bot's server → to the transcription service's processing pipeline → to their storage systems. At every hop, your meeting data exists on infrastructure you don't control. The bot isn't just a participant in your meeting — it's a pipeline that extracts your audio and sends it to the cloud.

This is true even if the service claims end-to-end encryption. The bot itself is the decryption point. It has to decode the audio to capture it, and from that point forward, the audio is in the service's hands.

There's a better way

The meeting bot architecture exists because cloud services need a way to capture audio without installing software on every participant's machine. But there's a simpler approach: record on the machine that's already in the meeting — yours.

Transcripted records meetings by capturing your Mac's system audio (what you hear) and your microphone input (what you say). It doesn't join the meeting. It doesn't appear in the attendee list. No one on the call knows it's running unless you tell them.

The audio is processed locally by Apple's Neural Engine. Nothing is uploaded. Nothing is sent to a server. The transcript is saved as a markdown file on your Mac. The entire pipeline — capture, transcription, speaker identification, file output — runs on your hardware.

What this means in practice

No awkward introductions. The client call starts normally. No bot joining, no "what's that?" conversation, no permission negotiation. You record your meetings the same way you'd take notes — privately and without disrupting the flow.

No enterprise blockers. Because nothing joins the call, there's nothing for IT to block. Transcripted works with Zoom, Google Meet, Microsoft Teams, Slack huddles — any app that produces audio on your Mac. It doesn't interact with the meeting platform at all.

No cloud exposure. The audio pipeline is entirely local. System audio → Neural Engine → markdown file. Your meeting data exists in exactly one place: your computer.

Better meetings. When there's no bot in the room, people are more natural. They say what they think. They don't perform for the recording. You get better conversations and, consequently, better transcripts.

The bot era is ending

Meeting bots were a necessary hack for an era when local processing wasn't powerful enough to handle real-time transcription. That era is over. Apple Silicon's Neural Engine can transcribe meeting audio locally, with speaker identification, faster than the time it takes to upload the audio to a cloud API.

The question isn't whether meeting bots will disappear. It's how long it takes for everyone to realize they don't need them anymore. Your meetings should feel like conversations, not surveillance. Transcripted records everything without anyone knowing — because it runs on your Mac, not in your meeting.