The short version: AI notetakers make you look organised and cost you deals, because the moment a prospect knows every word is being transcribed, they stop saying the thing that gets them to buy. I ran this as a real test on my own sales calls, not a theory, and I've gone back to scrappy handwritten notes plus a private post-call voice memo to AI. Here's the exact workflow and the numbers behind it.
The call that made me stop
Last spring I was on a discovery call with the owner of a mid-sized events company, forty employees, decent revenue, exactly the kind of client I like working with on AI strategy. I had Fireflies running in the background, the little bot joins the Zoom, says "recording started," sits there in the participant list with its blank little avatar.
Twelve minutes in, she said something that changed the whole tone. "Can you turn that off? I want to tell you what's going wrong here and I don't want it written down anywhere." I turned it off. What followed was fifteen minutes of the real problem: her ops manager was quietly job-hunting, two of her best people were burnt out covering for him, and she was terrified of what would happen if she brought in outside help and it leaked internally. None of that was in the transcript. All of it was the reason she signed with me three weeks later at £4,200 a month.
That's the moment I started paying attention to what a recording bot does to a room, and it's not what the sales tools tell you it does.
What's really happening when the bot joins the call
Every AI notetaker company will tell you the same thing: never miss a detail, focus on the conversation, let AI handle the notes. All true. What none of them mention is that people change what they say the second they know a machine is capturing it word for word.
This isn't a hunch. Watch what happens in your own calls once you notice it:
- People stop naming names. Instead of "our finance director is the blocker," you get "there are some internal considerations."
- They stop admitting doubt. "I'm not sure this is even the right priority for us" becomes "we're evaluating our options."
- Numbers get vaguer. Real budget figures turn into "somewhere in that ballpark, depends on the board."
- They ask more clarifying questions upfront and reveal less, because they know the transcript could be forwarded, quoted back at them, or fed into someone else's AI summary tool later.
Every one of those things is the exact information you need to close the deal, and it's the exact information a recording makes people withhold. The bot doesn't lie to you. It just gets a more polished, more careful, less useful version of the person you're talking to.
The bit nobody in the AI sales tools space wants to say out loud
Here's the uncomfortable truth the notetaker companies won't put in their marketing: the same efficiency that makes AI notetaking brilliant for internal team meetings makes it actively worse for one-to-one sales and consulting calls, because those calls run on trust and disclosure, not on accurate record-keeping. A weekly ops standup benefits from a perfect transcript. A discovery call where you're trying to find out why someone hasn't fixed their broken marketing for three years does not. Those are different jobs and the software treats them as the same job.
The tools are optimised for the metric that's easy to sell, "never miss a detail," and not for the metric that pays your bills, "get the prospect to tell you the truth." I've watched fellow consultants proudly show off their Gong or Fathom dashboards full of talk-time ratios and sentiment scores while their close rate quietly sits at half what it used to be. Nobody connects the two because the dashboard looks impressive and the missing revenue doesn't show up as a line item anywhere.
The actual numbers from my own test
I'm not going to pretend this was a controlled scientific trial, I run a consultancy, not a lab. But I did track it across 2025 because I wanted to know if I was imagining the pattern.
- 40 discovery calls with an AI notetaker visibly running: 11 became paying clients. That's 27.5%.
- The next 34 calls with no bot, just me and a notebook, followed by a private voice memo to myself straight after: 15 became paying clients. That's 44%.
- Average deal size didn't change much, roughly £2,800 to £5,000 a month depending on scope, so this wasn't a case of the bot filtering in smaller, worse-fit prospects.
- Time spent on admin after each call went up by about eight minutes without the bot, since I have to dictate my own summary instead of copying one out. Worth every second of it.
Sixteen and a half points of close rate is not noise. That's the difference between a decent quarter and a brilliant one.
What I do instead, step by step
I still use AI heavily, I'm not anti-tool, I just moved it to a different part of the process where it can't spook anyone.
- No bot on the call, ever, for a first or second conversation. I take handwritten notes, badly, on an A5 pad. The scribbling itself signals I'm paying attention. People relax around a pen in a way they don't around a screen.
- Straight after hanging up, I record myself a two to three minute voice memo talking through what they said, in their words where I can remember them, plus my honest read on whether this is a real deal or a tyre-kicker.
- That memo goes into a transcription and summary tool, I use a simple app for this, and it turns into a structured note in my CRM: pain, budget signal, decision-maker, objection, next step.
- Before the call ever happens, I try to get as much of the boring detail as possible answered without a human conversation at all. I use short qualifying forms and, on my website, interactive tools people can play with on their own time, similar to the approach I wrote about in how interactive calculators convert more visitors, so the call itself is freed up for the honest, messy conversation instead of me asking "so what's your team size" for the tenth time that week.
- Only once a deal is signed and we're into project delivery do I turn recording back on for kickoff calls and reviews, because at that point the trust is established and everyone benefits from an accurate record.
When an AI notetaker is the right call
I want to be fair here because the tools aren't the villain, the wrong use case is.
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.
- Internal team meetings, standups, project reviews: brilliant, use it, stop taking your own notes.
- Client onboarding and delivery calls once the relationship is live and trust exists: fine, often preferred, clients like having a record of what was agreed.
- Support and troubleshooting calls where accuracy matters more than candour: fine.
- Any call where you legally must keep a record, financial advice, certain regulated services: mandatory, not optional, and in the UK you're required under GDPR to tell people plainly that the call is being recorded and roughly why, not bury it in a bot's auto-announcement they might not even hear.
- First-time sales and discovery conversations, negotiation calls, anything where you're trying to surface the real problem someone hasn't admitted to themselves yet: leave the bot off.
The trust cost is bigger than the transcript benefit
I think about this the same way I think about brands that win on personal trust rather than polish. Look at how Airbnb built a brand people trust enough to hand their house keys to strangers, it wasn't through more surveillance or more data capture, it was through removing friction and making people feel safe enough to be honest about what they needed. Every recording device on a call adds a tiny bit of friction back in, even when nobody says a word about it. Most prospects won't ask you to turn it off like my events company client did. Most will just quietly give you less.
The same logic runs through good follow-up email systems, the kind I've broken down when looking at how Mailchimp built a brand around trustworthy, useful communication, and it's the same logic Alex Hormozi hammers on constantly about closing deals: people buy from people they trust enough to be honest with, and anything that makes someone perform rather than talk kills that. I wrote about this in my piece on the business lessons Alex Hormozi practises, and the through-line is the same one I found by accident on that Zoom call: candour closes deals, polish doesn't.
If you're building this into your own sales process
Small business owners ask me for a one-line rule and here it is: record the calls where you need accuracy, and go bare-notes on the calls where you need honesty. If you can't tell which is which for your business, or you want someone to map your actual sales and client workflow rather than just bolt an AI tool onto whatever you're already doing, that's the kind of practical audit I run through my AI consultant work with small businesses. It's usually a two-hour conversation and it saves months of quietly losing deals to a well-meaning bit of software.
Free resource: The Discovery-to-Close Checklist.
Frequently asked questions
Do I have to tell clients I'm using an AI notetaker on a call?
In the UK, yes, plainly, under GDPR you need to make it clear recording is happening and roughly why, not just rely on the bot's automated announcement, which many people miss or ignore. In the US requirements vary by state, some require consent from all parties. Always disclose it out loud yourself at the start.
Does turning off the notetaker mean I lose accuracy in my notes?
Slightly, yes, but the trade-off favours you on sales calls. A voice memo recorded within minutes of hanging up, while everything is fresh, captures the useful signal, pain points, budget, objections, without the chilling effect of a live recording, and it's accurate enough for CRM purposes.
Which AI notetakers are worth using at all?
Fireflies, Otter.ai, Fathom and Gong are all solid for internal meetings, project reviews, and delivery calls once a client relationship is established. None of them are wrong tools, they're just the wrong tool for a first discovery conversation where you need someone to be uncomfortable and honest.
Will this improve my close rate or was that just one consultant's experience?
It's one business's numbers, not a scientific study, so treat it as a strong signal rather than a guarantee. The mechanism behind it, that people disclose less when they know they're being recorded, is well established in basic psychology and worth testing on your own calls for even one month before you decide either way.
Related reading: The AI Notetaker Sitting In On Your Client Calls Might Be Breaking Your NDA and AI Meeting Notetakers Are Quietly Costing You Client Trust.
