Is AI leaving your prospects cold?
Why letting an LLM write your prospect emails is quietly costing you, and what we actually use it for instead
We get a lot of approaches. Most days our inbox has a fresh handful of them, and we’re sorry to say we don't entertain them. Not because we’re precious, and not because we don't appreciate someone having a go at winning our attention. It's because we can tell, usually within a line and a half, that nobody actually wrote the thing. A machine did, and the person who sent it couldn't be bothered to read it back properly before it landed on us.
That's the bit that gets us. Not the AI. The can't-be-bothered.
So before anyone accuses us of being a luddite, let us be clear. We use AI every single day. It has become one of the most useful tools in our new business kit, and we'll explain how a bit further down. This isn't an anti-AI piece. It's a piece about one specific use of it that we think is both ineffective and, actually, a little bit rude.
The trap is that it feels brilliant
Here's the seductive part, and we've felt it too. You paste your brief into the box, you hit go, and back comes something that looks the part. It's articulate. It's on brief. It's well structured. You read it and think, that would have taken me forty minutes, and it took forty seconds. This is incredible!
And that little hit of satisfaction is exactly the problem.
Because you are not the only person getting that hit. Everyone is. Every other agency, every other business development person, every other hopeful chasing the same wish-list account you are. They've all pasted a near-identical brief into a near-identical box and received a near-identical answer back, dressed up to look bespoke. The thing that felt like an edge is now just the wallpaper. And nobody reads the wallpaper.
Polished used to be the goal. Now it's a tell.
This is the part that should make you sit up, because it isn't a matter of taste, it's showing up in the numbers.
Cold email reply rates have been falling for years. The benchmarks put it somewhere around eight and a half percent back in 2019, down to roughly five percent in 2025, and sliding towards three to five percent as we march through 2026. One of the named culprits, openly, is the sheer volume of AI-generated outreach now flooding inboxes. We have collectively spammed our own well dry.
But here's the genuinely interesting bit. People who test this stuff at scale are finding that the smooth, optimised, perfectly-flowing AI email now performs worse than a shorter, rougher, more direct one. The polish has become a negative signal. Buyers have developed what one writer nicely called a "delete reflex", a subconscious filter that bins anything carrying the familiar pattern before they've even consciously clocked why. "I hope this email finds you well" is practically a self-destruct button at this point.
So when you send the slick AI draft, you're not just being a bit lazy. You're actively triggering the thing that gets you deleted. Worst of both worlds.
The giveaways your prospect feels but can't name
Most people on the receiving end couldn't write you a list of why an email smells of AI. They just feel it and move on. But the patterns are well documented by now, and once you see them you can't unsee them:
Everything comes in threes. Neat little trios of adjectives, all perfectly balanced, often with a bit of alliteration, the way a consultancy names its values. Humans use threes too, but we break the rhythm and pick the wrong third word on purpose. AI does it every time, flawlessly, which is the giveaway.
The "it's not just X, it's Y" move. Sounds profound. Commits to absolutely nothing.
Sentences that are too well behaved. They never trail off. They never interrupt themselves. They never go on a small tangent and come back. Real writing has friction in it. This has none, and the absence is what you feel.
The vocabulary. Leverage. Robust. Unlock. Elevate. Underscore. Seamless. A whole dialect of words nobody says out loud at the pub. ("Delve" was the famous one, though it's gone a bit quiet lately.)
The throat-clearing openers. "In today's fast-paced world." "In an era of." Words that fill space while you wait for a point that often never properly arrives.
The em dash. Sprinkled everywhere. A perfectly good bit of punctuation, ruined by association.
A fair caveat, because we don't want to be the people who declare every em dash a robot. Some of these have been so widely flagged that real humans now nervously edit them out of their own writing, and plenty of people just write that way naturally.
Would you let an intern email your dream client?
Here's the question we keep coming back to.
Picture your top three wish-list accounts. The ones you'd genuinely love to win. Now imagine handing them to an intern on their first week and saying, "right, off you go, email them, win the business." Of course you wouldn't. Not because the intern is stupid, but because they don't yet have the judgement to know what good looks like, or what would land badly, or what to leave out.
That is more or less what's happening when we let AI write the email unsupervised. Here's the uncomfortable truth though. We are the interns. Many of us are deeply experienced in our fields, with years of experience, but we are genuine beginners at using these tools well. We don't yet have the instinct for what the machine gets subtly wrong, where it bluffs, where it sounds like everyone else. So we hand our best prospects to a tool we can't yet supervise properly, and we press send on its first draft. It's the intern emailing the intern.
What we actually use it for, and the one rule we never break
Now, the honest part. What AI is genuinely brilliant at, is the homework. We use it to research a prospect company inside out. To get up to speed on a sector we’re walking into. To understand the person we want to talk to, what they care about, what they've been building, where our world and theirs might actually connect. The legwork that used to eat a whole afternoon now takes twenty minutes, and that means more time we can spend on the part that matters: working out what we actually want to say, and then saying it as us.
But we have one rule, and we never break it. Think, do, think. We think about what we need before we ask. We let the tool do its bit. Then we think again and review everything it gave us for truth and for evidence before it goes anywhere near a prospect. AI is a confident liar. It will hand you a plausible fact that's plainly wrong with breezy certainty. In front of a prospect, that's not a typo, that's your credibility.
We're all flying without a licence
The bigger picture is this. A genuinely powerful technology got handed to everybody, all at once, with no training and no rulebook. Most of us are using it on instinct and crossed fingers. Mistakes are easy to make, and in new business a mistake in front of a decision maker is costly because you rarely get a second go at that introduction.
Which is why we think deliberate, human writing isn't nostalgia. It's the obvious competitive response to a market drowning in sameness. When everyone else sounds like the same machine, the person who sounds like a person stands out.
The least we can do
If we’re asking a company to trust us with their business, the very least we owe them is to have written to them ourselves. In our own voice. Having actually thought about them. That's not me being noble. It's us respecting their time and their inbox, and frankly it's us backing ourselves to be more interesting than a template.
Think before you paste.
