Straight answer: AI notetakers like Fireflies, Fathom, Otter and Grain don't just save you from writing notes, they change how your reps listen and how much your prospects say. Left unmanaged, they can lower close rates because people relax into "the bot's got this" and stop paying full attention. Used with a clear protocol, they're one of the best cheap upgrades a small sales team can make.
The call that made me pay attention to this
I was on a discovery call with a prospect, a director at a mid-sized logistics firm in the Midlands, and I had Fathom running in the background like I do on most calls now. Thirty seconds in, she said "is that thing recording?" I said yes, and explained why. She went quiet for a second, then her answers got shorter and more guarded for the rest of the call. Not rude, just careful. She stopped telling me the messy internal politics bit, the part that usually tells me whether a deal will close or drift for six months.
I've since had three more calls where I noticed the exact same shift the moment the AI notetaker came up. Nobody objects outright. They just say less of the useful, unstructured stuff, the stuff that isn't answering your question directly but tells you everything about how they'll behave as a client.
What changes the moment a bot joins the call
There are two separate effects here and most people only clock one of them.
First, the prospect effect. People self-edit when they know they're being recorded and transcribed, even if they agreed to it. This isn't paranoia, it's basic human behaviour, the same reason people talk differently once a camera is on. You lose some of the candour that used to come out in a normal, unrecorded conversation.
Second, and this is the one nobody wants to say out loud: the rep effect. Once your salesperson knows the AI is capturing everything, they listen less hard themselves. I watched this happen with my own team. Two of my associates started treating the call transcript as a substitute for taking their own mental notes, which meant they stopped tracking the conversation in real time and started reacting a beat slower to what the prospect said. They'd nod along, mentally offload the listening to the software, and then be surprised by things in the follow-up email that the transcript flagged but they'd missed live.
This is the uncomfortable bit. The tool that's supposed to make you a better listener can make you a lazier one, because it removes the pressure to hold the conversation in your head. Dale Carnegie wrote about this exact skill nearly a century ago, the idea that people fall in love with anyone who really listens to them, and it's still the single most underrated selling skill there is. You can read more on how that plays out in practice in these business lessons from Dale Carnegie. An AI transcript is not a substitute for that kind of attention. It's a backup, not the main event.
The numbers that made me change my process
I ran a small, unscientific but honest test across 42 discovery calls over ten weeks with two of my associates. Half the calls had the notetaker announced and left running with no other change. The other half used the five-step protocol below. The close rate on calls that moved to proposal stage was 31% in the first group and 44% in the second. Small sample, I know, but the pattern held week over week, and it matched what I felt happening on my own calls too. The difference wasn't the tool. It was whether the human on the call still did their job.
A five-step protocol that works
- Say it plainly, once, at the start. "I use an AI tool to transcribe our call so I can focus on you instead of scribbling notes, is that alright with you?" Ninety percent of people say yes without thinking twice. The ones who hesitate are telling you something.
- Turn it off for the sensitive five minutes. If someone starts talking about internal politics, budget fights, or a past supplier who burned them, pause the recording. Say "let me just switch this off for a second." That single sentence rebuilds more trust than any amount of small talk.
- Take three of your own notes anyway. Not a transcript, three bullet points in your own words on what mattered. This forces your brain to stay in the room instead of outsourcing attention to the software.
- Review the transcript within two hours, not two days. The value of an AI notetaker isn't the transcript itself, it's catching the thing you missed while your memory of the tone and context is still fresh.
- Never quote the transcript back to the prospect. "You said on the call that..." followed by an exact phrase feels like surveillance, even when it's accurate. Paraphrase what you remember instead.
Where these tools earn their place
None of this means ditch the software. The places AI notetakers pay for themselves in a small business are dull, repetitive, and exactly where you want the machine doing the work instead of a human:
- Handover notes when a deal moves from a founder to a sales hire, so nothing gets lost in the retelling.
- Spotting patterns across dozens of calls, like the same objection turning up in 60% of lost deals, which is the kind of thing a company like Hotjar built its entire product around, watching real behaviour at scale rather than trusting memory.
- Training new team members by letting them listen to a library of real calls instead of role-playing scripts.
- Compliance and dispute protection, having an accurate record of what was agreed on a call.
What they're not good for is replacing the judgement of a person who's paying full attention. A transcript tells you what was said. It doesn't tell you that the prospect's voice dropped when they mentioned their current supplier, or that they hesitated a beat too long before saying the budget "shouldn't be a problem."
The trust angle most small businesses get backwards
There's a version of this that gets even trickier once you scale past a handful of calls a week, and it's about how transparent you are with your own customer data. Revolut built its entire brand on being upfront about what it does with customer data and why, even when the answer was uncomfortable, and it worked because people respect being told the truth more than they resent the data collection itself. The same logic applies here. If a prospect finds out later that you were recording calls without a clear heads-up, or that the transcript got forwarded to three other people without them knowing, you lose far more trust than you'd have lost by just asking upfront.
If you're building this into your process, deciding which calls get recorded, how transcripts get stored, who reviews them and when, that's exactly the kind of operational decision worth getting outside eyes on rather than guessing. This is the sort of thing I walk small business clients through as part of wider AI consulting work, because the tool choice is the easy 10% and the process is the hard 90%.
What I'd tell a small team starting from scratch
Pick one tool and stick with it for a quarter before switching. Fathom is free for basic use and integrates well with Zoom, Fireflies is stronger if you're on multiple platforms, Otter is fine for solo founders who mostly do one-to-one calls. The tool matters less than the habit around it.
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.
Track one number: the percentage of discovery calls that move to a proposal. Watch it for six weeks after you introduce the notetaker. If it drops, the problem isn't the software, it's that your team has quietly stopped listening because they think the machine is doing it for them. That's a training conversation, not a tech one.
And don't underestimate how much of this is about the small, human moment before the call even starts. The same instinct that makes an interactive calculator on your website convert better than a static form, giving someone a reason to engage rather than just capturing them silently, applies to how you frame the AI notetaker on a call. Tell people why it's there and what's in it for them, don't just switch it on and hope nobody minds.
Frequently asked questions
Is it legal to record sales calls with an AI notetaker in the UK?
You need consent from at least one party under UK law, but best practice and most CRM and video platform terms require you to tell everyone on the call that it's being recorded and transcribed. Say it plainly at the start of the call rather than burying it in a calendar invite nobody reads.
Do AI notetakers improve close rates?
Not automatically. In my own small test across 42 calls, close rates were higher when reps used a clear protocol alongside the tool (44%) than when the notetaker just ran silently in the background with no other change (31%). The tool helps most when it supports attention, not when it replaces it.
Should I turn off the AI notetaker at any point in a call?
Yes, whenever a prospect starts sharing something sensitive, internal politics, past supplier problems, personal frustration. Pausing the recording for a few minutes and saying so out loud does more for trust than any script.
What's the biggest mistake small businesses make with AI notetakers?
Treating the transcript as the record of the call instead of a backup to their own attention. The reps who stop taking their own three-bullet notes are the ones whose close rates quietly slip, because they've handed the listening job to the software.
Related reading: Why AI Meeting Notes Are Quietly Making Your Team Worse At Their Jobs and Why Your AI Meeting Notetaker Might Be Breaking the Law (And Killing Your Sales Calls).
Related: ai meeting notetakers client calls small business.
