A sales rep I spoke with last month told me about a deal she lost because of a meeting tool. She was on a discovery call with a VP of operations at a logistics company. Two minutes in, a bot called “AI Notetaker” appeared in the participant list. The VP paused, asked what it was, and spent the next five minutes asking about data retention policies instead of talking about his supply chain problems. By the time they got back on track, the energy was gone. He ended the call early. She never got a second meeting.
She didn’t lose the deal because her product was wrong. She lost it because the tool she used to capture the meeting changed the meeting itself.
That story keeps coming up. Not always that dramatic, but the same pattern: the recording tool alters the thing it’s supposed to record. People speak differently when they know a bot is watching. They hold back. And the notes you get afterward are a faithful record of a worse conversation.
The blind spot in every meeting tool
The entire AI note taker category was built around one idea: capture what happened after the meeting ends, then package it. Transcript, summary, action items. Every product in the space does some version of this, and some do it well.
But there’s a question almost nobody in the category is asking: what if the meeting itself could be better?
Think about how most calls go. Someone’s on a noisy street. Another person’s accent is hard to follow over a choppy connection. The meeting happens, it’s mediocre, and then an AI produces a tidy summary of that mediocre conversation. The notes look clean. The conversation they captured wasn’t.
This is the gap that Krisp spotted. But to understand why it’s the company that filled it, you have to know where it came from.
A noise cancellation company that kept pulling the thread
Krisp didn’t start as an AI note taker. It started as an audio company, known for years as the app that removed background noise from calls. The technology worked at the device level, sitting between your microphone and whatever app you were using. Companies like Discord, Twilio, and OpenAI licensed it. Over 200 million devices and over a billion minutes of audio are processed monthly.
That’s a lot of meetings. And when you’re inside the audio stream of that many conversations, you start seeing things note-taking companies never see. Bad audio cascading into bad transcripts. One misheard word changing a summary, which changes an action item, which sends someone down the wrong path the next day.
So Krisp kept pulling the thread. If we’re already cleaning the audio, why not transcribe it? If we’re transcribing it, why not summarize it? If we’re summarizing it, why not structure the output around templates, push it to CRMs, make it searchable?
The result is a product that doesn’t separate “meeting quality” from “meeting notes.” They’re the same pipeline. Noise cancellation cleans up both sides of the call. Accent conversion, a feature nobody else in the space offers, makes speakers easier to understand in real time without changing their voice. By the time audio reaches the transcription engine, it’s already cleaner than what any competitor is working with. The notes are more accurate because the conversation was clearer, and the conversation was clearer because the audio was fixed before anyone started talking.
What this looks like in a real meeting
Back to that sales rep. She switched to Krisp after losing the deal. On her next discovery call, she selected Krisp as her microphone and started the meeting. No bot appeared. No notification popped up on the other side. She was calling from an airport lounge, and the background announcements and crowd noise vanished from her end. The prospect, calling from a busy open office in Mumbai, came through clearly too thanks to the two-sided noise cancellation and accent conversion.
When the call ended, she opened the app and found a structured summary: key points from the conversation, three action items with owners identified, and a full transcript organized by speaker. She’d set a “discovery call” template beforehand, so the notes were already formatted around the prospect’s pain points, budget, and timeline. The whole thing synced to HubSpot automatically.
No time spent writing a recap email. No second-guessing what was said. No awkward bot moment.
In sales, the follow-up you send within an hour of the call is worth more than the one you send the next morning after re-listening to a recording. That speed is where deals move forward or stall.
The part people don’t expect
There’s an AI chat feature that lets you query your past meetings like a database. “What pricing concerns came up in last week’s calls?” or “What did the engineering team decide about the migration timeline?” You ask, and it pulls the answer from the relevant moments.
This sounds minor until you’re the person taking six calls a day and someone asks on Friday what was decided on Monday.
Krisp also handles in-person conversations through its mobile app and can send a bot to attend meetings when you can’t be there. But the default is invisible. No bot, no extra participant, nobody knowing it’s running.
Why this matters beyond one company
The meeting tools market spent three years optimizing for the same thing: better summaries of the same bad meetings. Better transcripts, better structure, smoother integrations. All focused on packaging the output.
Nobody addressed the input. The call quality. The human experience of being in the meeting before any AI touches it.
Krisp’s bet is that fixing the meeting itself produces better everything downstream. And because the company spent years as an audio processing company before it became a meeting intelligence company, it had the technology to do it instead of just talking about it.
That’s the difference between a note taker that records your meetings and a tool that improves them. One gives you a transcript. The other changes what’s worth transcribing.
