Guide
You have seen it: a new client call starts, and an extra participant named something like "Notetaker AI" is sitting in the attendee list. Someone asks what it is. Someone else asks if the call is being recorded. The first two minutes of your most important meeting just got spent on your tooling. There is a better architecture — and the best version of it never uploads your audio at all.
The Problem
Most AI notetakers — Otter's OtterPilot and Fireflies.ai among the best known — work by sending a bot into your call as a participant. These are capable, mature products with strong transcription and integration ecosystems. But the bot model itself carries structural problems that no amount of polish fixes.
An unexplained participant on a sales call, a negotiation, or a sensitive HR conversation changes the room. People speak differently when a recorder is visibly present — usually more guarded, which is the opposite of what you want.
Many organizations restrict unknown participants, and some hosts simply deny the bot at the waiting room. When that happens, your note-taking system fails at the exact moment you needed it.
A bot needs a meeting link to join. Conference rooms, client visits, interviews, and site walkthroughs do not have one. For bot-based tools, the entire world of in-person conversation simply does not exist.
A bot means your conversation is being captured by a third-party service and processed on its servers. For some counterparties — and some industries — that is a conversation-ender before the conversation starts.
The Architecture
A bot-free meeting tool does not join the meeting — it listens from your side. On a virtual call, it captures the system audio your computer is already playing. In a room, it uses your device's microphone. Nothing appears in the participant list, nothing needs an invite, and nothing can be blocked at the door. Granola helped popularize this approach on the Mac, capturing audio locally instead of sending a bot — though per its public documentation, its transcription and AI processing still rely on cloud services. Which raises the real question: once the audio is captured, where does it go?
No extra participant in the call. Solves the optics and blocking problems. Audio may still be sent to cloud services for processing.
Audio is captured on your device rather than by a server-side bot. Better — but capture and processing are different things.
Audio is captured and transcribed on the device itself. Nothing is uploaded, ever. This is where Meeting Intelligence lives.
The Bot-Free Option That Goes Further
Meeting Intelligence is a native Swift app for iPhone, iPad, and Mac that transcribes entirely on-device using Apple's Speech framework. It works alongside Zoom, Teams, Google Meet, Webex, Slack Huddles, and FaceTime — and just as well with a phone on a conference room table. And because it is live during the meeting, it does more than take notes: Ask This surfaces strategic questions in real time, Quick Context explains acronyms and references the moment they come up, Watch For flags risks as they form, and Decision Intelligence splits every summary into Decisions Made and Decisions Needed. Meeting Memory carries context across meetings so the whole system gets sharper over time.
Use the managed Pro cloud tier for zero setup, run analysis fully locally with MLX on Apple Silicon Macs, bring your own Anthropic or OpenAI key, or point it at Ollama. The transcript itself is always produced on-device.
Transcript-only mode is free with no subscription. Pro is $12.99 per month or $99.99 per year, with a 7-day free trial and no payment required to start. Built by Edgepoint Solutions LLC.
Get Started
If a bot-based workflow fits your team, tools like Otter and Fireflies are solid choices. If you want AI meeting notes with nothing joining your calls and nothing leaving your device, download Meeting Intelligence and see the difference on your very next meeting.
Try Pro free for 7 days. No payment required to start. Transcript-only mode remains free.