Straight answer: if an AI notetaker joins your client calls without a clear heads-up, you're building your business on a small trust problem that gets bigger every week. It's not illegal in most cases, but it's shabby, it breaches UK GDPR's fairness principle more often than people realise, and it's the kind of thing that quietly costs you referrals. Fix the consent bit and the tool is useful. Skip it and you're gambling with client trust for the sake of not having to type your own notes.
What happens when the bot joins the call
Here's the scene. I'm on a discovery call with a prospective client, a director of a mid-size manufacturing firm in the Midlands. Ten minutes in, she stops mid-sentence and says, "Sorry, who's Fathom?" Fathom was the AI notetaker that had silently joined the Zoom call as a participant, sitting there in the little grid of faces like an uninvited guest. I hadn't told her it would be there. I'd assumed the calendar invite footer covered it. It didn't, not clearly enough for her to notice, and she felt recorded without being asked.
We lost about ten minutes of that call rebuilding trust. She wanted to know where the transcript went, whether it trained a model, whether I'd delete it. Fair questions. I didn't have great answers on the spot, which was the actual problem, not the bot itself.
That happened to me once. It's happened to thousands of small business owners this year and they don't even know it, because most people on the other end of the call are too polite to stop and ask who Fathom, Otter, Fireflies or Grain is.
Silence isn't consent under UK GDPR
This is the bit most AI productivity content skips entirely, because it's not exciting and it doesn't sell a tool. Under UK GDPR, recording someone's voice and processing it (which a transcript absolutely is) requires a lawful basis, and it needs to be done fairly and transparently. A line buried in a calendar invite that nobody reads is not meaningful transparency. Neither is a small pop-up message that flashes for two seconds when the meeting starts.
The Information Commissioner's Office, the UK's data protection regulator, is clear that people need to know recording is happening before or as it starts, not discover it by accident partway through. You can read the ICO's own guidance on this at ico.org.uk, and it's worth ten minutes of your time if you use any of these tools professionally.
Most small business owners aren't breaking the law on purpose. They're breaking it by default, because the tool's out-of-the-box setting is "join silently and record everything," and nobody goes into the admin panel to change it. That's the uncomfortable bit. The tool is set up to make consent your problem, not its problem, and almost nobody fixes it.
The bit that should worry you more than the legal risk
Here's what nobody's saying loudly enough. Even if you nail consent, most of these notetaker tools default to using your call data to "improve the product," which is company language for training their models on your client's voice, your client's words, your client's business problems. Client confidentiality clauses in your contracts probably don't cover this scenario at all, because whoever wrote your contract in 2019 wasn't thinking about AI listening in.
I've had clients tell me things on calls they would never put in an email. Cash flow worries. Staff they're about to let go. A rival they're worried is undercutting them. None of that was meant for a server somewhere being used to fine-tune someone else's product. This is the trade-off nobody frames when they're selling you the fifteen hours a week you'll save on note-taking. You're not just saving time. You're moving your client's private business problems into a system you don't control, on terms you probably never read.
This is where a lot of AI advice online quietly avoids the point, because "turn off the training toggle" doesn't make for an exciting headline. But it's the single most important click most small business owners haven't made.
Fix it before your next call: five steps
None of this takes longer than an afternoon.
- Turn off auto-join. Go into your Zoom or Teams admin settings and disable the notetaker joining every meeting automatically. Make it opt-in, per call, so you're deciding each time.
- Say it out loud. At the start of every call, one sentence: "Quick heads-up, I'm using an AI tool to take notes on this call, that alright with you?" It takes four seconds and it changes everything about how the relationship starts.
- Check the training toggle. Every major notetaker (Otter, Fireflies, Grain, Fathom) has a setting, usually buried in privacy or data settings, that lets you opt out of your data being used to train their models. Find it. Switch it off.
- Set a retention window. Don't let transcripts sit forever. Thirty to sixty days is plenty for most client work, then let them auto-delete.
- Put a line in your contract. One sentence covering how you handle call recordings and transcripts. Your client will trust you more for having thought about it, not less.
That's it. Five things, one afternoon, and you've turned a liability into a feature you can mention in your onboarding without flinching.
When these tools are worth using
I'm not anti-notetaker. I use one most weeks. The value is real: you stop half-listening while scribbling notes, you get an actual searchable record of what was agreed, and you can send a client a clean summary within minutes of the call ending, which looks sharp and professional. Done, with consent handled up front, it's one of the better small, boring uses of AI in a service business. It's not glamorous like a chatbot or an AI-written campaign, but it removes a genuine daily friction point.
The transcripts themselves are also more useful than most people realise once you stop treating them as disposable. A good chunk of my blog content and case study material now starts life as a rough line pulled from a client call transcript, the client's own words about a problem they had, cleaned up and turned into something useful. It's the same instinct behind how Ahrefs built trust with transparent, specific content instead of vague marketing speak: real words from real conversations beat anything you'd invent from scratch.
Where this goes wrong isn't the technology. It's businesses treating "I have an AI tool now" as the whole strategy, the same mistake plenty of brands make when they bolt AI onto marketing without thinking about trust first. Mailchimp's whole brand is built on trust and consistency, not on flashy tools, and that's the bit small businesses keep skipping when they rush to automate the human parts of client relationships.
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 wider pattern this is part of
This isn't really a story about notetakers. It's a story about small businesses adopting AI tools at the interface layer, the bit clients and customers see and feel, without thinking about consent, trust, or what happens to the data afterwards. Same pattern shows up with AI chatbots on websites that don't disclose they're bots, AI-written testimonials that never happened, AI-generated "case studies" with numbers nobody checked.
Alex Hormozi talks a lot about building businesses on things that compound, not shortcuts that erode trust for a short-term win, and one of the clearest business lessons from Hormozi's approach is that the boring, unglamorous fix (tell people what's happening, get their yes) almost always beats the flashy tool applied carelessly. Airbnb built an entire multi-billion pound business on trust between strangers, and Airbnb's marketing strategy leaned hard into transparency precisely because trust is the product when you're asking someone to hand over something personal, whether that's their spare room or their business worries on a call.
If you're already deep enough into AI tools that you're not sure what else in your stack has a consent or data problem you haven't spotted, that's worth a proper look rather than a guess. This is exactly the kind of gap an AI consultant for small business audit tends to catch quickly, because it's rarely the flashy tool that trips people up, it's the quiet default setting nobody checked.
What I do differently now
Every call I run now starts with the one-line heads-up. I've had exactly one person say no, and I simply took notes by hand for that call, which was fine. Everyone else says yes without a second thought, because being asked feels respectful even when the answer would have been yes anyway. That's the whole point. It was never really about the tool. It was about being asked.
My data retention is set to thirty days. My training toggle is off on every tool I use. It took me about forty minutes total across three tools to check and fix, and I did it after that call with the Midlands client, sat there a bit red-faced, going through settings menus I should have checked eighteen months earlier.
Frequently asked questions
Do I legally have to tell people an AI notetaker is on the call?
In the UK, yes, in effect. UK GDPR requires processing of personal data, which includes a voice recording and transcript, to be fair and transparent, and a hidden line in a calendar invite footer doesn't meet that bar. Say it out loud at the start of the call, it takes seconds and removes any ambiguity.
Which AI notetaker is safest for client confidentiality?
There's no single "safest" one, they all collect and store transcripts on their own servers. What matters more is your settings: turn off any option that lets your calls be used to train the company's models, set a short retention window, and check whether the tool offers a business or enterprise tier with stricter data handling, since free tiers usually have the loosest defaults.
Should I put AI notetaker use in my client contract?
Yes. One clear sentence covering that calls may be transcribed by an AI tool, how long you keep the transcript, and that clients can opt out for any specific call. It protects you, and it signals to the client that you've thought about their privacy rather than bolting on a tool because it was convenient.
Is it worth using AI notetakers at all given the risk?
Yes, once consent and data settings are sorted. The time saved on manual note-taking and the quality of the follow-up summaries you can send clients is real and worth having. The risk isn't the tool itself, it's using it on default settings without telling anyone, which is a fixable habit rather than a reason to abandon it.
Related reading: AI Tool Costs Explained for Business Owners in 2026 and AI Meeting Note-Takers: What Nobody Tells You Before You Turn One On.
For the bigger picture, see my full guide to AI marketing.
