AI-Written Outreach That Doesn't Sound AI-Written: The Rules We Use
TL;DR
AI can write outreach that gets replies, but only if you use it to be more specific to the reader, not to send more generic messages faster. The trick is letting AI do the research and the first draft, then having a human cut the filler, add one real detail, and keep it short. AI should augment your judgement, never replace it.
Why does most AI outreach get ignored?
Because it sounds like AI wrote it, and everyone can smell it now. You know the emails. "I hope this message finds you well. I came across your company and was thoroughly impressed by your innovative work in the industry." Polished, padded, and saying absolutely nothing. The reader's thumb is already moving to archive.
The problem isn't the AI. It's how people use it. Most folks ask the tool to "write a cold email to a roofing company" and send whatever comes back. The result is grammatically perfect and completely forgettable. It reads like it could have been sent to ten thousand other businesses, because it could have been.
AI outreach that converts does the opposite. It feels like one person took two minutes to think about one other person. That's the whole game. And the good news is AI is brilliant at the bit that makes that possible, the research and the first draft, as long as a human stays in charge of the judgement.
What's the difference between AI doing the work and AI doing the thinking?
AI should do the work; you should do the thinking. That one line solves most of the problem.
The work is the slow, repetitive stuff. Reading a prospect's website. Pulling out what they actually sell and who they sell to. Drafting three versions of an opening line. Tracking who replied and who went quiet. A human doing all that by hand is the reason outreach never gets done.
The thinking is the bit machines are bad at. Which angle will land with this particular person? Is this observation genuinely interesting or just filler dressed up as flattery? Does this sentence sound like something I'd actually say out loud? That's your job, and it takes seconds once the draft is in front of you.
When I'm doing outreach for the property side or for Darra Tyres, I never start from a blank page and I never send the first thing the AI hands me. I let it draft, then I attack it. Most of my editing is deleting.
See where AI fits in your business. Free.
A 45-minute audit. We map the highest-value automations and what they're worth in time and money. No pitch, no pressure.
How do you make an AI-written email sound human?
Cut, shorten, and add one specific thing. In that order.
First, cut the throat-clearing. Any sentence that exists to "warm up" the reader should go. "I hope you're well." "I wanted to reach out." "I'll keep this brief." Just be brief. Real busy people open with the point.
Second, shorten everything. AI loves long, balanced sentences with two clauses joined by a comma. People don't talk like that. Break them up. A four-line email outperforms a fourteen-line one nearly every time, because the long one looks like work and gets deferred forever.
Third, and this is the one rule that does the heavy lifting, add a single specific detail that proves you looked. Not "I love what you're doing." Something like "Saw you've just opened a second depot in Ipswich, staffing two sites must be stretching your booking system." That one line tells the reader a human was actually paying attention. AI can find that detail for you in seconds by reading their site. You decide which detail is worth mentioning.
If you only change one habit, make it this: ban the generic compliment, demand the specific observation.
What rules do we actually use?
Here are the ones we apply to every message, in plain terms.
Write to one person, not a segment. Even when AI is helping at scale, every email should read as if it was meant for that single reader. If you could swap in any other company name and it still works, it's too generic.
Lead with them, not you. The first two sentences should be about their world. Your offer comes later, and shorter than you think. Nobody cares what you do until they believe you understand what they do.
Plain words only. No "leverage", no "solutions", no "synergies". If you wouldn't say it leaning on a counter with a coffee, don't type it. AI defaults to corporate vocabulary, so this is where most of your editing goes.
One clear ask. End with a single, low-effort question. "Worth a quick call next week?" beats a three-option calendar link with conditions. Make saying yes easy.
Vary the follow-ups. This is where AI earns its keep. Most replies come from the second, third, or fourth touch, and those almost never get sent by hand. AI can track who's gone quiet and draft a fresh, short nudge each time, so the chase actually happens instead of dying in your to-do list.
Where does the human stay in the loop?
Right before anything leaves the building. That's the non-negotiable.
We don't believe in fully automated outreach that fires without a person seeing it. Not because the AI can't write, it can, but because the cost of one tone-deaf message to the wrong person is high, and the cost of a quick human glance is almost nothing. AI is there to augment your team's reach, not to take their name off the message.
In practice that means the system does ninety per cent of the labour: research, drafting, sequencing, tracking. A person spends a few minutes approving, tweaking a line, or killing a message that doesn't sit right. You get the volume of automation with the judgement of a human. That's the combination that actually books meetings.
The owners who get burned by AI outreach are the ones who removed the human entirely. The ones who win keep the human exactly where they add the most value, on the final read.
What does good look like in the end?
A short, plain email that one person could have written to one other person in five minutes, except the research and the draft and the follow-up tracking all happened in seconds, by machine. The reader can't tell, and frankly it doesn't matter, because the message is genuinely relevant and genuinely human in tone.
That's AI outreach that converts: not more messages, better-aimed ones. Not faster spam, faster relevance. The tool handles the grind; you handle the meaning. Get that split right and your reply rate climbs without you spending your evenings in your inbox.
If your outreach is either eating your time or sounding like a robot, or both, that's exactly the kind of thing worth a proper look. Book a free AI audit with Anaboo and we'll walk through where AI can quietly take the grind off your team while keeping the human voice front and centre. No pressure, no jargon, just a practical conversation.
Live with passion & AI,
Brett
Host a podcast? Have Brett on as a guest.
Straight talk on implementing AI in real SMEs, no jargon, plenty of receipts from the businesses we run.
Frequently asked questions
Will prospects be able to tell I used AI to write the email?
+
Not if you use AI for the heavy lifting and a human for the judgement, AI drafts and researches at speed, but you set the angle, cut the filler, and add the one specific detail that proves a person was paying attention.
Isn't AI outreach just spam at scale?
+
It becomes spam when you automate the sending without automating the relevance, the rule is to use AI to make each message more specific to the reader, not to fire more generic messages faster.
How much should I personalise each message?
+
Enough to prove you looked, usually one genuine, specific observation about their business near the top, with the rest of the message kept short, plain, and about them rather than you.
What's the biggest giveaway that an email was written by AI?
+
Polished, padded language that says nothing, phrases like "I hope this email finds you well" and grand claims with no specifics; real people write shorter, plainer, and more pointed.
Can AI handle the follow-ups too?
+
Yes, and that's where most of the wins are, AI can track who hasn't replied and draft timely, varied follow-ups so the chase actually happens, while you approve before anything goes out.
Brett is a four-time founder (Darra Tyres, Gladfish, EzyTrac, Anaboo) and the operator behind AIOS, Anaboo's AI Operating System. He writes from inside the build, installing AI in his own businesses first and reporting back what actually moves the numbers. Based between Singapore, the UK and Australia.
