Straight answer: AI meeting notetakers like Otter, Fireflies and Fathom save real time, but they carry two risks almost nobody mentions: they can leak sensitive information to the wrong person with one wrong click, and they quietly make you a worse listener in your own client calls. Use them, but build in the checks below before you let one anywhere near a sales or client meeting.
Next step on this topic: How An AI Meeting Assistant Saves You Time On Note-Taking (And Where T.
What happened when I tried this
I ran Fireflies on client calls for three months last year. It joins your Zoom or Teams meeting as a little robot participant, transcribes everything, and emails a summary to whoever you tell it to.
It worked well for about ten weeks. Then it sent a call summary to the wrong Slack channel. The summary included a discounted retainer rate I’d agreed with one client, and that channel was shared with a second client who was paying full price for the same package. I found out when the second client emailed me asking, quite politely, why he was paying more.
He didn’t cancel on the spot. He renegotiated, and I lost around £4,200 a year off that contract, plus a chunk of trust I had to work to earn back over the next six months. The tool didn’t do anything wrong technically. I had it set to auto-share to a workspace integration I’d forgotten was still connected from an old project. One setting, one click I never made deliberately, one expensive mistake.
That’s the bit that never makes it into the “AI tools for small business” listicles. These tools are built to share information as widely and quickly as possible, because that’s what makes them look impressive in a demo. Nobody in the demo shows you the settings page where old integrations quietly stay connected.
The GDPR problem you’re probably ignoring
If you run any of these tools on a call with someone in the UK, you’re recording and processing their personal data, and under UK GDPR and the Data Protection Act 2018 you generally need to tell them and get their agreement, not just mention it in a privacy policy nobody reads. The Information Commissioner’s Office has guidance on this, and the short version is: silence is not consent. Saying “just so you know, this call is being transcribed by an AI tool, is that alright with you” takes eight seconds and covers you. Most people never bother.
The uncomfortable truth about what these tools do to your listening
Here’s the part I didn’t expect. Once I knew a bot was capturing every word, I stopped listening as carefully. Not consciously. I just relaxed, the way you relax into a satnav and stop reading road signs.
The problem is that client calls aren’t just data. A client who goes quiet for two seconds before answering a pricing question is telling you something. A joke that lands flat is telling you something. Fireflies transcribes the words perfectly and misses all of it, and if you’re relying on the summary instead of your own attention, so do you.
I noticed this in a call with a prospect who kept saying “it’s fine, whatever works” about timelines, in a tone that clearly meant it was not fine. The transcript recorded “client confirmed timeline is acceptable.” I nearly sent that summary straight to my calendar and moved on. I only caught it because I happened to remember the call myself. If I’d been leaning on the bot the way I’d started to by week eight, I’d have missed a client who was two weeks from walking.
This is the trade nobody talks about honestly. You get your time back and you lose a bit of your edge, and the edge is usually the thing that keeps clients.
How to use one of these without it biting you
I still use an AI notetaker. I just run it differently now.
- Say out loud, every single time, that the call is being transcribed, and wait for a yes before you carry on.
- Check your sharing settings before every new client relationship, not once when you first installed the tool. Integrations drift.
- Read the summary yourself before it goes anywhere near a client’s inbox. Five minutes, every time, no exceptions.
- Turn the bot off for the first call with any new prospect. That first call is where you’re reading tone, hesitation and body language, not collecting a transcript.
- Set a deletion schedule. Fireflies and Otter both let you auto-delete recordings after a set period. Thirty days is sensible for most small businesses. There is no good reason to be sitting on eighteen months of client audio.
- Never let the bot’s summary replace your own three-line note on what the client cares about. The bot captures words. You capture what matters.
What these tools cost
Otter’s free plan gives you 300 transcription minutes a month, which is roughly ten hour-long calls, fine for a solo consultant with a light call load. Fireflies’ free tier is similar, around 800 minutes total storage rather than monthly. Once you’re doing real client work, you’ll likely need a paid tier: Fireflies Pro runs about $18 per user per month, Otter’s Business plan is around $20 per user per month, and Fathom is free for the core transcription and summary but charges for team features and CRM syncing. Against that, if a notetaker saves you even 30 minutes of manual note-writing after every call and you’re doing eight calls a week, that’s roughly four hours back weekly, which at a consultant’s typical rate of £75 to £150 an hour is worth far more than the subscription. The maths works. The risk sits elsewhere.
Where this fits with the rest of your marketing
If you’re already trying to build trust with prospects before they ever get on a call with you, an AI notetaker mishap is exactly the kind of thing that undoes months of work. The businesses that handle this well tend to be the same ones doing the basics everywhere else, the kind of consistent, slightly boring trust-building that companies like Airbnb built their entire early growth on, long before anyone talked about AI on calls at all, as I’ve written about in this breakdown of Airbnb’s marketing strategy.
The follow-up email your AI notetaker generates also matters more than people think. A robotic, bullet-pointed AI summary landing in a client’s inbox two minutes after a warm, human call can feel jarring. If you’re using a tool like Mailchimp to nurture that relationship afterwards, it’s worth rewriting the AI summary into your own voice before it becomes part of that sequence. I go into the mechanics of building that kind of nurture flow in my piece on Mailchimp’s marketing strategy.
There’s also a broader point here that Alex Hormozi makes well, even though he’s talking about sales volume rather than AI: speed and volume without a quality check just means you make your mistakes faster and more often. That’s the whole risk of AI notetakers in one line, and I unpack the wider principle in these business lessons from Alex Hormozi.
If you’d rather capture information from prospects without recording a live conversation at all, an interactive calculator on your site does a similar job earlier in the funnel, no consent conversation needed, and it gives you structured data instead of a transcript you still have to read.
When it’s worth it and when to skip it
Use an AI notetaker if you’re doing five or more client or discovery calls a week and you’re currently writing notes from memory an hour later, because that’s where the real time saving shows up. Skip it for sensitive negotiations, HR conversations, anything involving pricing that differs between clients, and first calls with new prospects where reading the room matters more than capturing every sentence.
And if your tech stack has grown to the point where you’re not sure which tools are still connected to which client folders, that’s usually a sign you need someone to look at the whole setup rather than one tool at a time. That’s exactly the kind of audit I do when I work with businesses as an AI consultant for small business, because the risk is almost never the AI tool itself, it’s the three integrations nobody remembers switching on.
None of this means the tools are bad. Fireflies has saved me hours over the past year. It just means the setup you leave running on autopilot is the part that costs you, not the transcript itself.
Frequently asked questions
Is it legal to use an AI notetaker on client calls in the UK?
Yes, but you generally need to tell the other person and get their agreement before you start recording or transcribing, under UK GDPR and the Data Protection Act 2018. A quick verbal heads-up at the start of the call is usually enough and takes seconds.
Which AI notetaker is best for a small business or solo consultant?
Fathom is a solid free starting point if you just need transcripts and summaries without much sharing complexity. Fireflies and Otter both offer stronger integrations and search across past calls once you’re on a paid plan, useful once you’re running more than a handful of calls a week.
Can an AI notetaker send a transcript to the wrong person?
Yes, and it happens through forgotten integrations more often than through the tool itself. If a summary is set to auto-share to a Slack channel, CRM, or shared folder from an old project, it will keep sharing there even after that project ends unless you check and remove it.
Does using an AI notetaker make client calls worse?
It can, if you rely on it instead of listening. Transcripts capture words accurately but miss tone, hesitation and sarcasm, all of which often matter more than the literal sentence a client used. Treat the AI summary as a backup to your own notes, not a replacement for paying attention.
Related reading: The AI Notetaker Sitting In On Your Client Calls Might Be Breaking Your NDA and AI Meeting Notetakers Are Quietly Costing You Client Trust.
