You're on a call with a potential buyer. They're telling you what actually matters: the budget, the two neighbourhoods they'd consider, the one deal-breaker ("no ground floor, not after last time"). And you're trying to listen and type at once, knowing you'll remember maybe half of it accurately by Friday. An AI notetaker offers to fix exactly that: join the call, transcribe everything, hand you a clean summary five minutes later. It's a real fix for a real problem. It's also, this year, one you can't just switch on without a step most guides skip entirely.
What these tools actually do
Tools like Fathom and Fireflies are AI meeting notetakers: they join your call as a visible participant, or run quietly in the background on your desktop, and produce a transcript, a summary, and a list of action items afterward. You can search a transcript later (for example, "did she mention a deadline?") and jump straight to the moment it was said. Both have real free tiers (Fathom: unlimited recording and transcription; Fireflies: 400 stored minutes), with paid plans landing somewhere between €10 and €20 a month once you want more. This isn't hype-cycle vapourware. It's a genuinely useful tool that a one-person estate agency, or any business that lives on phone consultations, can afford today.
Where it gets it wrong
It doesn't always know who said what: two voices on a call routinely get merged or swapped in the transcript. It struggles with anything outside everyday vocabulary: a street name, a mortgage product, an acronym your trade uses without thinking, a client's surname. Every serious review of these tools says the same thing: you edit the transcript afterward, you don't publish it raw. Treat the AI's version as a strong first draft of what was said, not a court record. For an estate agent, that means checking the numbers it caught (a budget, a square-metre figure) before you act on them, not trusting the summary blind.
The part almost every guide skips
Here's what the product pages leave out: recording a client's voice and turning it into searchable text is processing their personal data, and that comes with rules: not someday, now. Spain's data protection authority (the AEPD) published guidance this April specifically about AI voice transcription, and it's stricter than most people assume. The headline point: joining a call is not consent. A blanket line buried in your email footer or your terms of service doesn't count either. Consent has to be specific to that call, given before it starts, and it lapses when the call ends. The guidance also expects you to signal that recording is happening for as long as it's happening, not just once at the start, and to actually check what your notetaker vendor does with the recording afterward (where it's stored, who can see it, whether it trains a model) before you pick a tool, not after a client asks.
In practice this is one sentence, said out loud, every time: "I'm using an AI tool to take notes on this call so I don't miss anything, is that alright?" That's it. Most clients say yes, because it's clearly for their benefit too. The few who hesitate have just told you something worth knowing before the call goes any further.
Honest limits
This doesn't belong in every conversation, and it shouldn't be forced into one. If a client is telling you something they clearly don't want written down (a health issue, a money problem, a family situation shaping their decision), turn the recorder off and take a note by hand, or don't note it at all. And if you genuinely can't find, in five minutes, where your chosen tool stores its recordings or whether it uses them for anything beyond your own account, that's not a detail to skip past. Pick a different tool, or don't record until you can answer it.
Start this week
Before your next client call, write the one-sentence disclosure above somewhere you'll actually see it, and use it: every call, not just the ones that feel sensitive. Then open your notetaker's settings once and find the data retention and storage answer, so you're not guessing the next time someone asks. If you're already careful about where your business's own data ends up, the same instinct applies here. I wrote about the training-data side of that question in Is Your Business Data Training Someone Else's AI?; this is the client-facing half of the same habit.
Clara F.