Straight answer: AI meeting notetakers like Otter, Fireflies and Fathom are brilliant for internal team calls but risky on client and prospect calls, because most people say things off the cuff that they never expect to see written down and shared. The fix isn't ditching the tools, it's asking consent out loud every time and turning them off for the parts of the conversation that matter.
The call that changed my mind
Three months ago I had a discovery call with a prospect worth roughly £3,500 a month in retainer, a nice bit of business for a small consultancy like mine. Fireflies was running in the background, as it does on almost every call I take now. Halfway through, the prospect got candid about her old agency, said some pretty unflattering things about a named competitor, then laughed and said "obviously don't repeat that."
Except I had. The transcript landed in her inbox forty minutes later as an automated summary, because Fireflies sends the recap to everyone on the call by default unless you switch that off first. She emailed me within the hour, polite but clearly rattled, asking me to delete it. I did, and I apologised, but the deal went quiet after that and never closed. That single default setting cost me a client.
That's the bit nobody tells you when they're raving about how AI notetakers save you two hours a week on admin. They do save the time. They also record the human moments people assume are private, and those moments are usually the ones that matter most in a sales or client conversation.
The numbers that get thrown around, and the ones that don't
Otter.ai's business plan runs about $20 a month per user, Fireflies is around $10 to $19 depending on tier, Fathom's paid plan sits near $19 a month, and there are free tiers on all three that cap you at a limited number of minutes. Vendors love quoting the time saved: a rough industry estimate is that manual note taking eats 10 to 15 minutes per hour-long call once you add writing it up and filing it. Multiply that by twenty calls a week and you can see why small businesses jumped on this fast.
What doesn't get quoted anywhere near as often is the trust cost. A 2024 Cisco consumer privacy survey found that 76% of people say they wouldn't buy from a company they don't trust with their data, and meeting transcripts are data too, just data people forget is being collected because it feels like a conversation, not a form they filled in. I've never seen a SaaS company put that stat next to their own signup page, for obvious reasons.
Why "record everything" quietly backfires
Here's the uncomfortable bit. Small business owners tend to think recording everything makes them look organised and professional. Mostly it makes the other person feel watched. There's a difference between a client knowing a call is being transcribed for internal notes and a client realising, mid-sentence, that every offhand comment is now a searchable document sitting in someone else's software.
GoPro built an entire brand on the idea that capturing everything is the whole point, and it works brilliantly for action sports because the person filming and the person being filmed are usually the same person, choosing what gets shown. Client calls are the opposite. One side is doing the recording, the other side often doesn't clock it's happening, and that gap is where the trust damage sits. If you want to see how a brand handles that kind of tension between capturing everything and respecting the audience, the GoPro marketing strategy is worth a read, because their whole positioning depends on consent being obvious, not buried in a settings menu.
The same logic runs through Hotjar's business, which literally sells tools that watch how people use a website. What's kept them credible for over a decade is that they built transparency into the product itself rather than bolting it on as a privacy policy nobody reads. There's a proper breakdown of how they did that in the Hotjar marketing strategy piece, and it's the exact playbook most small businesses using AI notetakers skip entirely.
The policy I use now
I didn't stop using AI notetakers. I stopped trusting the defaults. Here's the process, step by step:
- Say it out loud at the start of every call. Not buried in a calendar invite. "This call is being transcribed by AI so I can focus on the conversation, is that alright with you?" Takes four seconds, gives the other person a real chance to say no.
- Turn off auto-send of the summary. In Fireflies this is under Notetaker settings, in Otter it's under sharing preferences. I now review every transcript myself before anything goes to the client, which takes maybe three minutes per call.
- Pause the recording for anything personal. Most tools let you pause and resume mid-call. If someone starts talking about a divorce, a health scare, or bad-mouthing a competitor, I pause it. That single habit would have saved the £3,500 deal.
- Delete the raw audio after 30 days. I set this as a calendar reminder because none of the three tools I've used do it automatically by default, they keep everything unless you go digging for the retention settings.
- Never let the AI summary go out as the final word. AI transcription still mishears names, numbers and prices a fair amount, so I always rewrite the summary in my own words before it reaches a client's inbox.
That's five extra minutes of admin per call. It's a small tax against a large risk.
Where AI notetakers earn their keep
Internal team calls are a different story entirely, and this is where I'd never go back to manual notes. A weekly ops call with a virtual assistant or a freelance designer, an internal strategy session, a training call where you want an accurate record of what was agreed, those are exactly what these tools were built for. Nobody's confessing anything they'll regret, and having a searchable transcript six weeks later when someone asks "did we agree on the November deadline or the December one" is worth every penny of the subscription.
I also use the automated follow-up angle a lot, similar to how IFTTT built its whole product around "if this happens, automatically do that" logic without anyone having to think about it again. There's a good look at how they turned that simple idea into a business in the IFTTT marketing strategy post, and I run a rough version of it myself: transcript gets filed, action items get pushed to my task list, and a calendar reminder pops up in seven days if a follow-up hasn't happened. That part of the automation is fine, because it's process, not people's private confessions.
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.
What to do if you're rolling this out across a small team
If you've got even two or three people on client-facing calls, don't just let everyone pick their own notetaker and settings. Write a one-page policy: which tool the team uses, what gets recorded, what gets paused, how long transcripts are kept, and who reviews summaries before they go external. It took me about ninety minutes to draft ours and it's saved several awkward conversations since.
If you'd rather not build that policy from scratch, or you're trying to work out where AI belongs in your client process versus where it's a liability, this is the kind of thing an AI implementation coach earns their fee sorting out, because most small business owners get the tool right and the governance wrong.
There's also a case for building your own trust into the front end of your process before anyone even gets to a call. A well built interactive calculator or intake form does some of the qualifying work up front, so the discovery call itself has fewer surprises and less need to record every word to "capture everything," because half of it's already on record in writing, submitted by the prospect themselves, with their full knowledge.
The habit that protects you
Duolingo built a business on tiny, repeatable habits rather than one big feature, and the AI notetaker discipline works the same way. It's not one big decision, it's a small habit repeated on every single call: say it out loud, pause when it gets personal, review before it's sent. Read through the Duolingo marketing strategy breakdown and you'll notice the pattern, small consistent actions beat one clever system every time, and that's exactly what keeps AI notetakers safe rather than risky.
And if you want the calmer version of all this, worth looking at how Calm built an entire brand around slowing down rather than capturing more. The Calm marketing strategy is a useful reminder that not every business advantage comes from recording, tracking or automating more. Sometimes the advantage is simply being the person who pauses the recording and says "this bit's just between us."
Frequently asked questions
Should small businesses use AI notetakers on sales calls?
Yes, but tell the prospect out loud that the call is being transcribed rather than relying on a line buried in a calendar invite, and turn off auto-sent summaries so you can review the transcript before anything reaches their inbox.
What's the biggest risk with AI meeting transcription tools?
The biggest risk isn't accuracy, it's that people say things off the cuff assuming they're private, and default settings on most tools automatically share the full transcript with everyone on the call within minutes.
How much do AI notetaker tools cost for a small business?
Otter's business plan is around $20 a month per user, Fireflies runs roughly $10 to $19 depending on the tier, and Fathom's paid plan sits near $19 a month, with limited free versions available on all three.
Is it legal to record a client call with AI without asking?
In the UK and most US states you're required to get consent before recording a call, so asking out loud at the start isn't just good manners, it's the legal baseline in most jurisdictions, and it's worth checking your specific state or country's rules before relying on silence as consent.
Related reading: Why Your AI Meeting Notetaker Might Be Breaking the Law (And Killing Your Sales Calls) and Business Lessons from Emily Weiss.
