Straight answer: If an AI note-taker joins your call and records or transcribes someone without telling them clearly beforehand, you are very likely breaching UK GDPR’s transparency rules, not just being a bit rude. Most small business owners treat this as a productivity hack. It is also a data protection decision, and almost nobody has written down a policy for it.
What’s showing up on your calls now
Eighteen months ago, the only thing joining a Zoom call was the person you’d invited. Now it’s them, plus a small parade of software: Otter.ai, Fireflies, Fathom, Read.ai, or Zoom’s own AI Companion, quietly transcribing, summarising, and in some cases pulling out “action items” and emailing them round afterwards.
I counted my last 20 client and podcast calls. Fourteen had at least one AI note-taker join automatically, uninvited by me, sometimes announced by a little bot name popping into the participant list halfway through the call. On three occasions, two different bots joined the same call, both owned by different attendees, both recording independently.
The tools themselves are cheap and good. Fireflies runs about £8 to £15 a user a month. Otter’s business plan is roughly £16. Fathom has a useful free tier. Read.ai sits around £15 to £20. For that money you get a searchable transcript, a summary, and often a highlight reel of who talked the most (which is its own uncomfortable conversation for another day). None of that is the problem. The problem is what almost nobody says out loud before pressing record.
The bit that gets skipped before anyone hits record
Under UK GDPR, a voice recording that can be linked to an identifiable person is personal data. Transcribing it, storing it, and letting an AI tool summarise it counts as processing that data. The law is fairly plain that people need to be told this is happening, what it’s for, and where it goes, before it starts, not somewhere in a footer nobody reads. The Information Commissioner’s Office has been clear for years that covert or unannounced recording of calls involving personal data sits on shaky ground, and the government’s own data protection guidance spells out the same transparency principle: people have a right to know when their data is being collected and why.
Here’s the uncomfortable part most advice on this topic skips over: the fact that a bot joins with a visible name in the participant list is not the same as informed consent. Most people on a call are half looking at their inbox. A bot icon appearing is not a request. It is a notice, and a weak one at that. If someone later says “I didn’t realise that was being recorded and kept,” that’s a real problem, not a technicality, especially if the call touched on pricing, a complaint, a personal circumstance, or anything that could end up in a dispute.
Why this bit further than it looks
A transcript is permanent and searchable in a way memory never was. I learned this the sharp way. On a call last spring I made an offhand, slightly grumpy comment about a mutual contact, the sort of thing you say and forget five minutes later. Six weeks after that call, a line from that exact transcript turned up, quoted almost word for word, in an email from someone I’d never spoken to. The client’s note-taker had generated a shareable summary, and it had been forwarded further than either of us expected.
Nothing came of it beyond an awkward apology on my part, but it changed how I think about every call I take now. It’s not paranoia to say that a throwaway line in a “casual” meeting can now live forever in a searchable database, get pulled into a summary, and land in an inbox with your name attached, weeks after you’d stopped thinking about it. Building genuine client trust the way Howard Schultz built Starbucks on consistent, human treatment of both staff and customers depends on people feeling they can speak freely with you. A silent recording bot chips away at that, even when nobody means any harm by it.
What I changed after that email landed
I didn’t stop using AI note-takers. I use Fireflies most weeks, and it’s saved me hours chasing up what was agreed on a call. What changed was the ten seconds before every call starts.
- I now say, out loud, on the recording: “This call is being transcribed by [tool name], I’ll share the summary with you afterwards, tell me now if you’d rather it wasn’t.” That single sentence covers most of the transparency requirement in one breath.
- I turned off the “auto-join” setting on my own note-taker so it never sneaks into a call I haven’t planned for it to join.
- I added one line to my calendar invite template: “Note: I use an AI assistant to transcribe calls for accuracy. Let me know if you’d prefer this off for our conversation.”
- For anything sensitive, contract negotiations, complaints, HR conversations, I switch the bot off entirely and take my own notes like it’s 2015.
- I stopped forwarding raw transcripts to third parties. If someone else needs the gist, they get my summary, not a machine-generated word-for-word record of someone else’s meeting.
None of that took more than an afternoon to set up, and it’s the sort of decision Napoleon Hill would have called making up your mind and sticking to it: the business lessons from Napoleon Hill that get quoted most often are all versions of the same idea, decide once, apply it every time, and stop relitigating it call by call.
A five-step policy any small business can put in place this week
If you run even a two or three person team, write this down rather than leaving it to individual judgement, because individual judgement is exactly how the awkward email situation happens.
- Pick one tool, not five. Every team member using a different note-taker means five different data retention policies and five different privacy notices nobody’s read. Standardise on one.
- Write one sentence and make everyone say it. “I’m recording and transcribing this with [tool] so I don’t miss anything, shout if that’s a problem.” Ten seconds, every call.
- Set a retention limit. Most tools let you auto-delete transcripts after 90 or 180 days. Do it. A transcript nobody needs anymore is just a liability sitting in a database.
- Decide what never gets recorded. Complaints, disciplinary chats, anything with a minor involved, anything a client flags as confidential. Put it in writing so it’s not a judgement call at 4pm on a Friday.
- Check where the data lives. Some tools store transcripts on US servers. If you’re a UK business handling UK client data, know what that means for your privacy notice, and update it if it’s out of date.
If putting any of this together with your team feels like more time than you have this month, an AI implementation coach can build a simple policy and staff briefing into an afternoon rather than it sitting on your to-do list for a year, which is where most good intentions like this go to die.
The upside nobody mentions either
It’s worth saying plainly: done, this is one of the best small business tools of the last few years. Client calls that used to disappear into vague memory are now searchable. I’ve pulled exact wording from three-month-old calls to settle “did we agree that” disputes in my own favour, cleanly, with no argument, because the transcript simply said what was said. Sales teams are mining their own call transcripts for the exact phrases prospects use, then feeding those phrases straight back into landing pages and ad copy, which is the same instinct behind how Ahrefs built its marketing strategy around real customer language rather than guessed keywords. If you’d rather give visitors something interactive to engage with instead of just recording every conversation they have with you, tools like the interactive calculators covered here do a similar job of gathering useful data, except the person filling it in knows exactly what they’re handing over, which is really the whole point.
Brands that get remembered for the right reasons, the way Away built its reputation on being straight with customers rather than clever with them, tend to be the ones that treat transparency as a feature rather than paperwork. An AI note-taker that everyone on the call knows about, agreed to, and can see the summary from afterwards is a genuine time-saver. One that shows up silently and gets forwarded around six weeks later is a small, entirely avoidable mess.
Frequently asked questions
Is it legal to use an AI note-taker on a client call without telling them?
Under UK GDPR, recording and transcribing someone’s voice counts as processing personal data, and the law requires transparency about that before it happens. Not telling the person is legally shaky, particularly if the call touches on pricing, complaints, or anything personal. Announce it at the start of every call rather than assuming a bot’s name in the participant list counts as consent.
Which AI meeting bots show up most often right now?
Otter.ai, Fireflies, Fathom, Read.ai, and Zoom’s built-in AI Companion are the ones most small businesses run into, either because their team uses one or because a client’s does. Pricing runs roughly £8 to £20 a user a month, with Fathom offering a usable free tier.
What should a small business AI meeting policy include?
One standard tool across the team, a spoken heads-up at the start of every recorded call, a fixed retention period (90 to 180 days is common), a clear list of conversations that never get recorded, and knowledge of where the transcripts are stored so your privacy notice matches reality.
Do I need consent from every person on a call, or just whoever booked the meeting?
Every identifiable person on the call whose voice is being recorded needs to be told, not just the person who scheduled the meeting. If a third party joins partway through, they need the same heads-up, which is why saying it out loud at the start of the call, every time, is safer than relying on a calendar invite nobody re-reads.
Related reading: The AI Meeting Notes Habit That’s Quietly Making Your Team Worse at Their Jobs and The AI Notetaker in Your Sales Calls Is Costing You More Than You Think.
