The short version: AI notetakers like Otter, Fireflies, Fathom and Read.ai save you real hours every week, but they also change the meeting itself, people say less, hedge more, and go quiet exactly when things get interesting. Use one, disclose it every single time, and know that the transcript is a legal document the moment it exists, not just a handy summary.
Worth reading next: The AI Meeting Notetaker Problem Nobody Warns You About.
I started recording every call in early 2025
I'd been taking my own notes in client meetings for about twenty years, half typed, half scribbled, always missing the bit that mattered because I was too busy nodding. So when Fireflies started showing up in my calendar invites automatically, I let it in. Then I added Fathom for the Zoom calls and kept Otter for anything recorded on my phone during in-person meetings. For about eight months I had a bot in nearly every conversation I had for work.
The time saving is real and I'll give you the number rather than the vague promise. Across a normal working week I was spending roughly 90 minutes writing up notes after calls. With a notetaker doing the transcript and the summary, that dropped to about 15 minutes of editing. That's five hours a month back, which at my day rate is not a small thing.
But somewhere around month four, something shifted that nobody warns you about before you start.
The meeting where someone asked me to turn it off
I was on a call with a client who runs a mid-size recruitment firm, going through a marketing plan that touched on some staffing changes she hadn't announced yet. Fathom was recording, as usual, with the little "recording" badge visible in the Zoom window. About twelve minutes in she stopped mid-sentence, looked at the badge, and said ", can we pause that for a second, this bit's off the record."
I turned it off. We had the conversation. It was the most useful ten minutes of the whole meeting, the actual thinking, the doubts, the "here's what I really can't say to my board yet." And none of it got captured, summarised, or fed into anything.
That's the bit almost nobody writing about AI meeting tools mentions: the transcript changes what people are willing to say, in real time, while it's happening. Not because they're hiding anything shady. Because most useful business conversation is exploratory, half formed, "let me think out loud here", and very few people think out loud in front of a permanent, searchable, forwardable record.
What happens to a transcript once it exists
This is the part I wish someone had spelled out for me before I got comfortable. Once a meeting is transcribed, it stops being a conversation and becomes a document. That document can be:
- Subpoenaed or requested in a legal dispute, exactly like an email would be
- Searched by anyone with admin access to your notetaker account, including a future hire you haven't met yet
- Quoted back at you word for word, months later, in a way your memory of the conversation never would be
- Shared automatically to a Slack channel or CRM by an integration you set up once and forgot about
None of that is a reason to avoid these tools. It's a reason to treat every recorded meeting the way you'd treat a written contract, because functionally, that's what it now is.
The consent problem people skip past
In the UK, under GDPR, you need to tell people you're recording and processing their data, and you need a lawful basis for keeping it. In practice that means saying, out loud, at the start of the call: "just so you know, I've got an AI notetaker running, it's transcribing this for my own records, shout if you'd rather I turned it off." That takes eight seconds. I now say it on every single call, including with people I've worked with for years, because the badge on the screen isn't consent, a verbal heads up is.
Most people say yes without thinking twice. About one in ten asks a follow-up question, usually "where does the data go" or "who else can see it." A small number, maybe one in twenty, ask you to turn it off for the whole call. Respect that instantly and don't make it weird.
What I'd tell a small business owner setting this up from scratch
If you're not technical and the idea of choosing between Otter, Fireflies, Fathom and Read.ai feels like one more decision you don't have capacity for, I wrote about exactly this kind of overwhelm in how I use AI day to day without being a techy person, and the same logic applies here: pick the one that's already built into a tool you use, not the one with the most features.
Here's the practical setup I'd recommend, in order:
Want AI doing the heavy lifting in your marketing?
I build the systems that handle the boring 80 percent, so you get your week back. Done properly, with the human kept in.
- Pick one notetaker and use it everywhere, don't mix three tools like I did, it makes your transcript library a mess to search later.
- Turn off any automatic sharing or CRM sync until you've reviewed a summary manually at least a dozen times, so you know what it gets wrong.
- Write yourself one sentence of disclosure and say it every time, word for word, so it becomes automatic rather than something you fumble.
- Set a personal rule for what never gets recorded, for me that's anything involving another person's redundancy, health, or finances.
- Delete transcripts you don't need after 90 days rather than letting them pile up as a liability with no upside.
If you're doing this across a team rather than just for yourself, that's a different job, someone needs to set the policy, pick the tool, and train people on what not to say near a live microphone, and that's the kind of thing worth getting outside help with rather than guessing, which is exactly the gap an AI implementation coach is there to close.
Where it earns its place
I'm not anti these tools, I still use one on almost every call. Where they've paid for themselves:
- Client onboarding calls, where someone gives you six pieces of information in forty minutes and you need all six, accurately, not your paraphrase of them
- Pitch calls for guest posts or partnerships, where the exact wording of what an editor said they wanted matters later, the same precision I talk about in getting a guest post accepted, editors remember exactly what they asked for and get annoyed when you misquote it back to them
- Internal strategy reviews, the kind where three months later someone says "did we agree that" and now you can just check
Where I've turned it off entirely: any conversation about someone's job security, any negotiation where I want room to float a number and take it back, and anything a client explicitly frames as "just thinking out loud with you."
The uncomfortable bit
Here's what I noticed reviewing eight months of my own transcripts back to back: roughly 40 percent of a typical client call is small talk, repetition, and people talking themselves round to a decision they'd already half made before the call started. The notetaker doesn't just save you time writing notes up, it shows you, in cold text, how much of your meeting time is filler. That's not a flattering thing to discover about your own calls, and it's the reason I now run shorter meetings than I used to. Not because AI told me to. Because it showed me the receipts.
It also showed me, uncomfortably, that I talk more than my clients do on calls I think of as "listening" calls. The transcript doesn't care what I thought the meeting was. It just tells you what happened, word count and all, the same unglamorous honesty I found going through my own site's numbers when I wrote about the month my website started paying me again, the data doesn't flatter you, it just tells you the truth and leaves you to decide what to do with it.
A quick note on the tools themselves
I'm not going to pretend one is definitively best because it depends entirely on what you're already using. Otter is the cheapest entry point at around 16.99 dollars a month for the Pro plan and works well as a standalone app. Fireflies integrates cleanly if you live in Google Meet or Zoom and has a generous free tier. Fathom is free for the core recording and summary feature and is the one I'd point a total beginner towards first. Read.ai adds sentiment and talk-time analysis on top, which is useful if you want the uncomfortable truth about who's dominating your calls, but it's more than most small business owners need on day one. If you want to see what else is out there that most people have never heard of, my list of 50 websites you didn't know existed has a few tools in a similar vein worth a look.
Frequently asked questions
Is it legal to use an AI notetaker without telling everyone on the call?
No, not in the UK. Under GDPR you need to disclose that you're recording and processing personal data and give people a chance to object. A visible "recording" badge on screen is not the same as verbal consent, so say it out loud at the start of every call.
Do AI meeting notetakers save time, or is that overstated?
They do save time, but the number is smaller than most marketing pages claim. In my own week to week use it cut post-meeting writeup from around 90 minutes to about 15 minutes, roughly five hours back a month, not the "get your whole day back" promise you'll see in ads.
Which AI notetaker should a small business start with?
If you're already on Zoom or Google Meet and want the simplest free option, start with Fathom. If you want a cheap standalone app that works across phone calls and in-person meetings too, Otter's Pro plan at around 16.99 dollars a month is the better fit.
What should never get recorded by an AI notetaker?
Anything involving another person's job security, health, finances, or unannounced business decisions. Set a personal rule before your first meeting, not after an awkward one, and turn the bot off without making it a big deal when someone asks.
Related reading: AI Notetakers Are Quietly Costing You Client Trust. Here's What I Do Instead and Why I Turned Off My AI Meeting Notetaker After 14 Months (And What It Was Really Costing Me).
