There's a specific moment that makes people search for this: a bot named "Otter.ai Notetaker" appears in the participant list of a client call, and someone asks what it is. Or IT blocks it. Or the client asks you to remove it, and now you're taking notes by hand in a call you thought was handled.
Bot-based transcription has been the default for years because it's the easy engineering answer — join the meeting as a participant, get a clean audio feed. But the model is under pressure from three directions at once, and there's a straightforward alternative that most people don't know exists: an AI note taker without a bot.
Why bots are getting harder to use
Platforms are actively blocking them. Microsoft Teams began auto-blocking external AI notetakers in August 2026. Zoom administrators can restrict third-party apps under Admin > Advanced > App Marketplace, switching to an allow-list so only approved apps connect at all. There's no single "block AI bots" toggle in Zoom — it's marketplace controls and meeting locks — but if your IT department has locked things down, the effect is the same: your bot doesn't get in.
Legal exposure is real, not theoretical. A wave of federal class actions is testing whether Otter, Fireflies, and Microsoft Teams have been violating wiretap statutes. Fireflies has faced allegations under the Illinois Biometric Information Privacy Act over collecting voiceprints without written consent. Twelve US states require all-party consent — California, Connecticut, Delaware, Florida, Illinois, Maryland, Massachusetts, Montana, New Hampshire, Oregon, Pennsylvania, and Washington — and a bot in the participant list is a very visible piece of evidence about what was recorded and when.
It's socially awkward. This is the underrated one. A visible third-party bot changes how people talk. Candid conversation gets less candid when there's a named recorder in the room, and in sales or partnership calls that costs you something.
The alternative: record from your own machine
A Mac can capture two audio streams locally — your microphone and the system audio coming out of your speakers. Together those are the whole conversation. No bot joins, nothing appears in the participant list, and the audio never leaves your machine.
This works identically across Zoom, Google Meet, Teams, Slack huddles, FaceTime, a phone call on speaker, or an in-person conversation, because it operates at the audio layer rather than through each platform's API. That's the part people find surprising: there's no per-platform integration to break.
Vext does this. Hit the meeting shortcut. It records both streams, and transcription runs on-device with Parakeet TDT v3 on Apple Silicon. Speaker labels are generated locally, and voice profiles persist across meetings — name someone once and future meetings recognize them. Summaries with key points and action items come from a local LLM. We walked through the mechanics in transcribing meetings on a Mac locally.
brew install muvon/tap/vext
The honest tradeoffs
This isn't strictly better than a bot. It's a different set of compromises, and you should know which ones you're accepting.
You have to be in the meeting. A bot can attend a call you skip and hand you the transcript. Local recording can't. If your workflow depends on transcribing meetings you don't attend, bots genuinely solve a problem local recording doesn't.
No shared team workspace. Bot services give you a searchable org-wide library with permissions, comments, and CRM sync. Local recording gives you files on your Mac. For a sales team that needs every call in HubSpot, that's a real gap, and no amount of privacy argument closes it.
Nobody gets automatic notice. This cuts both ways. A bot's visible presence is an awkward disclosure mechanism, but it is a disclosure mechanism. Recording locally means the obligation to tell people moves entirely onto you — and in all-party consent states that obligation is legal, not just polite. Say "I'm recording this for my notes" at the top of the call. The lack of a bot doesn't lower the bar; it just means the bar is yours to clear.
One machine, one perspective. You capture what your Mac hears. If your own connection drops audio, that's what lands in the transcript.
What you get back
Nothing uploads. For NDA work, legal calls, medical conversations, HR discussions, or anything involving unreleased product, this is the entire argument. There's no vendor sub-processor list to review, no cross-border transfer question, no breach surface holding your meeting audio. The compliance conversation gets dramatically shorter when the answer is "the audio never left the laptop."
No indefinite retention by default. Many notetaker tools default to keeping everything forever, building a persistent searchable database of every meeting you've had. Local files are yours to delete.
No per-seat subscription. Otter Pro is $16.99/month billed monthly. Fireflies Pro is $10/seat/month annually. Granola Business is $14/user/month. Vext is $49 once — why we price it that way.
It works when the platform doesn't. In-person meetings, phone calls, a conference room conversation, a platform your bot vendor doesn't support. Audio is audio.
Which to choose
Use a bot if you need transcripts of meetings you don't attend, or if a shared searchable team library with CRM integration is central to how your team works. Those are real requirements and local recording doesn't meet them. We compared the options in Otter.ai alternatives for Mac.
Record locally if your calls involve confidential material, if your clients or IT department object to bots, if you're in an all-party consent state and want the recording footprint minimized, or if you'd rather not pay a subscription forever for something your Mac can already do.
Most people who switch don't do it for privacy in the abstract. They do it because a bot got blocked, or a client asked them to remove it, and they needed something that just works on the next call. If you want the fuller argument about what these tools do with your data, that's AI meeting notetakers and privacy.