Bottom line: AI notetaker bots like Otter, Fireflies and Fathom will save you an hour of typing after every call, but they also change how clients see you, sometimes for the worse, and almost nobody selling you the software mentions that part. Use them, but with rules you set yourself, not the defaults the app ships with.
The call that made me switch my settings
Eighteen months ago a prospect joined a discovery call, saw “Fireflies.ai Notetaker has joined the meeting” pop up in the participant list before I’d even said hello, and typed in the chat: “sorry, I need to leave, didn’t realise this was being recorded.” Gone. No callback, no email reply. Twelve years running my own business and that was the fastest I’ve ever lost a lead, before I’d asked a single question.
That was the moment I stopped treating AI notetakers as a neutral admin tool and started treating them as part of my client experience, same as my email signature or how fast I reply to a DM.
What these bots are doing on your calls
Otter.ai, Fireflies, Fathom, Read.ai and the built-in transcription now baked into Zoom and Teams all do roughly the same job: they join or listen, transcribe every word, then generate a summary with action items. Fathom’s free tier is good enough for most solo consultants. Fireflies Pro runs around 18 dollars a month per seat. Otter Business is closer to 30 dollars a seat if you want team-wide search across transcripts.
What the pricing pages don’t tell you is where that transcript goes afterwards. Most of these tools store recordings and text on their own servers, often outside the UK, for months by default. If a client mentions a number they’d never put in writing, a legal issue, a health condition, a redundancy they haven’t announced yet, it’s now sitting verbatim on a third-party server with a retention setting you probably never checked.
The uncomfortable bit: transcripts make you sound worse, not better
Here’s the part almost nobody selling these tools admits. A polished human summary of a call makes you look sharp. A word-for-word transcript makes every “um,” every half-finished sentence, every moment you talked over the client, permanently visible. I’ve had a client forward me my own transcript with a line highlighted: “you said the project would take six weeks here, but your proposal says eight.” I hadn’t lied, I’d just thought out loud on the call the way everyone does. The AI didn’t know the difference between a firm commitment and me thinking aloud, and neither did the client once it was in black and white.
Transcripts are unforgiving in a way memory never is. Before these tools, a slightly wrong number said mid-call got smoothed over in the follow-up email. Now it’s evidence.
The generic-summary trap
The second problem is quieter but costs you more over time: every AI-generated follow-up email starts to sound the same. “Great chatting today, here’s a summary of what we covered, next steps below.” Clients who work with several consultants or agencies get three or four of these a week, all in the same flat AI voice, and they stop reading them. I noticed my own reply rates on post-call emails drop once I started sending Fireflies’ auto-summary instead of writing my own three lines on top of it.
The businesses that win attention on a crowded channel are the ones that resist the template, the same lesson I keep coming back to when I look at how GoPro built a brand people notice rather than blending into the feed. A follow-up email is a brand touchpoint too. If it reads like it came out of a machine, that’s because it did, and clients can tell.
What I do now, step by step
- 1. Say it out loud before the bot joins. “I use an AI notetaker so I can focus on the conversation instead of typing, is that alright with you?” Takes four seconds. Nobody has ever said no to my face, but plenty would have quietly left if I hadn’t asked.
- 2. Turn off auto-join for first calls. First discovery calls are where people are most guarded. I take my own notes for call one, and only bring the bot in from the second meeting once there’s some trust built.
- 3. Set retention to the shortest option the tool allows. Fireflies and Otter both let you shorten storage windows in settings. I keep mine at 30 days. If I need something longer, I save the specific bit I need, not the whole recording.
- 4. Never let the raw transcript go to the client unedited. I read it, pull out the three things that matter, and write those in my own words. Takes ten minutes, saves the relationship.
- 5. Redact anything financial or personal before it hits my CRM. Numbers, names of other suppliers, anything a client said that wasn’t for the record, gets removed before the summary gets filed.
The data question nobody asks until it’s too late
If you’re in the UK, client conversations recorded and processed by a US-based AI tool sit in a grey area under UK GDPR, depending on where the servers are and what the vendor’s data processing agreement says. Most small business owners never read that agreement. I hadn’t, until a client’s compliance team asked me to send it over, which is a document you want to have read before someone asks for it, not after. If you’re running this across a team rather than just yourself, this is exactly the kind of policy gap an AI implementation coach earns their fee sorting out in an afternoon, because getting it wrong once with the wrong client can cost you the account.
Where these bots earn their place
None of this means don’t use them. For internal meetings, team check-ins, or calls where everyone already knows they’re being recorded, they’re brilliant. I run a weekly content planning call and Fathom’s summary saves me a solid 40 minutes of manual notes every single week, roughly 35 hours a year I’ve got back for actual client work. For repeat clients who’ve been through a few calls with me and know exactly how I operate, the bot joins without a second thought because the trust is already there.
The mistake is treating every call the same way. A cold prospect, an existing client of three years, and an internal team stand-up are three completely different trust levels, and your notetaker settings should change with them, not stay on the same default forever.
Alternatives worth trying if you want the notes without the discomfort
If the bot joining a call still feels off for certain clients, there are lighter-touch ways to get the same outcome. A short interactive quiz or interactive calculator sent before the call gathers the information you’d normally spend the first fifteen minutes asking for, so there’s less to transcribe and less pressure on the call itself. Some consultants I know share a Figma board live during the call instead of recording it, which does the same job of capturing decisions visually without a third-party bot listening in, similar to how Figma built collaboration into the product itself rather than bolting it on afterwards.
The follow-up matters more than the transcript
What moves a deal forward isn’t the AI summary itself, it’s what you do with it in the next 24 hours. A well-timed follow-up email, written by you, referencing something specific the client said, beats a perfectly formatted AI summary sent on autopilot every time. Joe Pulizzi’s whole approach to content and client relationships, which I’ve written about in these business lessons from Joe Pulizzi, comes back to the same point again and again: consistency and genuine attention beat volume and automation. An AI notetaker should feed your judgment, not replace it.
The same logic applies if you’re building a follow-up sequence off the back of these calls. I still run mine through Mailchimp, and the strategy lessons in how Mailchimp built its email marketing approach apply just as much to a single follow-up as a full campaign: personal beats polished, every time.
What I’d tell a small business owner starting this today
Pick one tool, not three. Fathom’s free tier or Fireflies at the entry price is plenty for a solo operator or a team under ten people. Set your retention window before your first call, not after your tenth. Announce the bot every single time until it becomes second nature to ask. And read at least one full transcript end to end in your first week, because seeing exactly how blunt and literal these tools are will change how you talk on calls faster than any article telling you to.
Free resource: The AI Weekly Planning Template.
Frequently asked questions
Are AI notetaker apps like Otter or Fireflies legal to use in client calls in the UK?
Generally yes, but you need consent from everyone on the call, which most UK guidance treats as a requirement under GDPR when personal data is being processed and stored. Announcing the bot before it joins and giving people the option to decline covers most small business use cases.
Should I tell clients I’m using an AI notetaker?
Yes, every time, even with repeat clients. It takes seconds to say and the alternative, someone noticing the bot’s name in the participant list and feeling ambushed, has cost me a lead before and will cost you one too.
What happens to the data these bots collect?
It’s stored on the vendor’s servers, often outside the UK, for a default period that’s usually longer than you’d choose yourself. Check your settings and shorten the retention window, and read the data processing agreement if you work with clients who care about compliance.
Is there a way to get meeting notes without an AI bot joining the call?
Yes, take your own notes on early or sensitive calls, use a pre-call form or short questionnaire to gather information in advance, or use collaborative tools like a shared whiteboard during the call instead of a listening bot.
Related reading: Why AI Pricing Advice Is Quietly Training Small Business Owners to Undercharge and Why Your AI Meeting Notetaker Might Be Breaking the Law (And Killing Your Sales Calls).
