Straight answer: AI meeting bots like Otter, Fireflies and Fathom are brilliant for internal calls but they're quietly damaging trust with clients who never agreed to be recorded, transcribed or fed into someone's training data. Ask permission before the bot joins, know exactly where the recording lives afterwards, and for anything involving money, contracts or a nervous client, turn the bot off and take your own notes. It costs you ten extra minutes. It's worth it.
Worth reading next: how to get client budget.
Worth reading next: AI Meeting Notetakers Are Quietly Costing You Client Trust.
Worth reading next: Your AI Meeting Notetaker Is Quietly Damaging Client Trust. Here's Wha.
The meeting that changed my mind
Eighteen months ago I was on a discovery call with a prospective client, a family-run manufacturing business in the Midlands, third generation, the sort of company that still does a lot of business on a handshake. Two minutes in, my Fireflies bot joined the Zoom call as it always does, showing up in the participant list as "Fireflies.ai Notetaker" with its little robot avatar.
The owner stopped mid-sentence. "Who's that? Is someone else listening?" I explained it was a transcription bot, that it just typed up notes so I didn't have to. He asked me to remove it. He then spent the next ten minutes visibly more guarded than he'd been in the first ninety seconds, and the call never really recovered. We didn't win that piece of work. I'm not saying the bot cost me the client on its own, but it changed the temperature of the room, and I noticed.
That's the bit nobody selling you these tools tells you: the discomfort is real, and it's not paranoia. A recording of a business conversation, especially one touching on pricing, cash flow or a family disagreement about the company's future, sitting on a third party server somewhere is a different thing to a human taking handwritten notes that live in a notebook on your desk.
What these tools do with the recording
I went back through the terms of service for the three tools I've used most, because I'd never read them before that call. Here's what I found, in plain English:
- Otter.ai free plan gives you 300 minutes a month of transcription. The Business plan is around $16.99 per user per month billed annually. Otter's terms allow them to use recordings to improve the service unless you're on an enterprise plan with a specific data protection addendum.
- Fireflies.ai runs $10 to $19 per user per month depending on tier. Recordings and transcripts are stored on their servers by default and can be shared to a team workspace automatically, which means a colleague you never intended to loop in can end up with access to a sensitive client conversation.
- Fathom is free for the core plan, which is exactly why it's grown so fast, but free products need to make money somewhere, and their terms include the right to use anonymised data to improve their models.
None of this is scandalous. It's standard software business practice. But standard software business practice and what your client thinks is happening to their words are two very different things, and that gap is where trust gets quietly eroded, meeting by meeting, without you noticing until a client says something like "who's that" and you realise you've never explained it.
The consent problem nobody wants to deal with
In the UK, under UK GDPR, recording a call where personal data is discussed generally requires you to tell people it's happening and, in most interpretations, get their agreement, not just announce it and carry on. Plenty of small business owners skip this because the bot joining automatically feels like a formality, not a decision. It isn't. If your client is in the US the rules vary by state, some are one-party consent, some are two-party, and if you work across borders like I do, between the UK, the US and clients in Israel, you're dealing with three different legal expectations in the same week without necessarily realising it.
I asked my own email list about this, a group of roughly 4,200 small business owners and marketers, a simple one-line poll: would you be uncomfortable if a supplier's AI bot joined a call with you without asking first. Just over six in ten said yes, uncomfortable, and around a quarter said they'd actively think less of the business afterwards. That's not a scientific study, it's a gut check from people who run real businesses, but it lines up with what I saw on that Midlands call. People don't mind AI. They mind not being asked.
What I do now instead
I haven't stopped using AI notetakers. I use them constantly for internal team calls, for my own thinking sessions, for webinars where the audience already knows they're being recorded. What changed is how I use them on client-facing calls, and it's a short list:
- I ask before the bot joins, not after. A line in the calendar invite: "I use an AI note-taking tool on calls, happy to turn it off if you'd rather I didn't, just say the word." Nobody has ever said no when it's framed as a choice rather than an ambush.
- For anything involving money, contracts, redundancy, or a family business decision, the bot stays off. I take my own notes, badly, by hand, the old way. It's slower. It's also honestly a better listening exercise, because typing while someone talks makes you a worse listener than you think.
- I checked where the data goes and turned off training use wherever there was a setting for it. Most of these tools have a toggle buried in account settings, usually under privacy or data preferences, that stops your recordings being used to improve the model. It's rarely on by default. Go and check yours today, it takes about four minutes.
- I delete recordings after the summary is done and shared, rather than letting them sit in the cloud indefinitely, which is the default on most plans.
None of this is complicated. It's the kind of unglamorous housekeeping I've written about before when I talked through how I use AI day to day without being technical, because the tools are easy to use badly if you never look at the settings, and most people never look at the settings.
When AI notetakers are still absolutely worth it
I want to be fair here because I'm not anti these tools, I use them most days. Internal team meetings, project planning calls, my own solo brainstorms where I talk through a problem out loud and want it transcribed so I can turn it into a blog post later, that's exactly what they're built for and they're brilliant at it. I've built entire content plans off a rambling forty-minute voice memo turned into a clean transcript in ninety seconds. The productivity gain is real and I wouldn't go back.
The line I now draw is simple: if the person on the other end of the call is a client, a prospect, or anyone whose livelihood or personal situation is the subject of the conversation, the bar for consent goes up. If it's colleagues who already know the score, the bar stays where it's always been.
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.
The uncomfortable maths on what silence costs
Here's the bit most posts on this topic skip entirely, because it's less flattering to the tools than "save five hours a week." If a client leaves a call feeling recorded rather than heard, and you lose that piece of work, the AI notetaker didn't save you time, it cost you the entire fee. I've seen consultants quote anywhere from £3,000 to £15,000 for the kind of project I was pitching that day in the Midlands. Ten minutes of asking permission first is a laughably cheap insurance policy against that. The tools that promise to save you time only save you time if the relationship survives the call, and that's the trade-off nobody selling AI notetaking software puts in their marketing.
It's the same lesson I learned the hard way rebuilding my own content strategy, the kind of thing I wrote about when I looked at what happened after publishing 569 AI-assisted blog posts and watching Google index just 5 percent of them. Speed and volume feel like progress right up until you check whether the thing you built holds up, and by then you've usually already spent the budget.
A quick script if you want to start using consent
If you want to fix this in your own business this week, here's exactly what to change, in order:
- Add one line to your calendar invite template mentioning the AI notetaker and offering to turn it off.
- Open your notetaker's settings and switch off "use my data to improve the product" or the equivalent, it's usually under Privacy or Data Preferences.
- Set an auto-delete rule for recordings after 30 days rather than letting them sit forever.
- For your next three client calls involving money or a sensitive decision, take your own notes instead and see how the conversation feels different. It does.
I found most of this simply by digging through settings menus and terms of service pages that nobody reads, the same way I stumbled across half the useful tools I now rely on, which is also why I keep a running list of the odd, under-the-radar sites and tools worth knowing about, the sort of thing I collected in 50 websites you didn't know existed but really should.
Rebuilding trust with clients after a bad experience takes far longer than losing it did, which is a lesson that runs through most of what I've relearned while rebuilding my own business publicly, including the month my own website started paying me again after I stopped chasing shortcuts and started paying attention to the boring details that build trust over time. AI notetakers are one of those boring details. Worth getting right.
Frequently asked questions
Is it legal to use an AI notetaker on client calls without telling them?
In the UK it sits on shaky ground under UK GDPR if personal data is discussed, since transparency and, in most readings, consent are expected before recording. In the US the rules depend on the state, some require agreement from everyone on the call, others only need one party to know. The safest approach for any small business is to always mention it upfront regardless of the strict legal minimum, because the trust cost of not asking is usually higher than the ten seconds it takes to ask.
Which AI notetaker has the best privacy settings for client work?
All three major tools, Otter, Fireflies and Fathom, offer a setting to opt out of your recordings being used to improve their models, but it's rarely switched on by default. Check your account's privacy or data preferences page directly rather than assuming, since the default setting varies by plan and changes as these companies update their terms.
Should small businesses stop using AI notetakers altogether?
No. They're useful for internal meetings, solo planning sessions and any call where everyone already understands and accepts the recording. The change worth making is narrower: raise the bar for consent specifically on client-facing calls involving money, contracts or anything personally sensitive, and take manual notes there instead.
How do I bring up AI note-taking with a client without it feeling awkward?
Put it in the calendar invite before the call, not as a surprise once you're both on the line. One sentence works: "I use an AI note-taking tool on calls, happy to turn it off if you'd rather I didn't." Framing it as their choice rather than your default removes almost all the awkwardness, and in my experience nobody has ever said no when asked that way.
Related reading: Why I Turned Off My AI Meeting Notetaker After 14 Months (And What It Was Really Costing Me) and Your AI Meeting Notes Are Making Your Team Worse Listeners (Here's What I Do Instead).
For the bigger picture, see my full guide to AI marketing.
