Six months ago, I'd have told you AI had no place in email marketing. Today, I use it every day. Not because I changed my mind. Because the tools got better, and I got honest about what they're good at and what they're terrible at.
The conversation around AI in marketing has been unhelpfully binary. On one side, the enthusiasts claiming it'll replace every copywriter alive. On the other, the sceptics insisting it produces nothing but soulless drivel. Both camps are wrong, and both are missing the practical middle ground where AI actually earns its keep.
Let's start with what works, because some of it genuinely surprised me.
Subject line generation. This is where AI punches well above its weight. Give it your email content, your audience segment, and your goal, and it'll produce 20 subject line variations in seconds. Not all of them will be good. Maybe five will be usable. But those five will include angles you wouldn't have thought of, and at least one will outperform what you'd have written on your own.
We tested this on a campaign for a managed print provider. Our human-written subject line ("Your print costs in Q3: a quick comparison") hit a 32% open rate. The AI-suggested alternative ("The number your facilities team hasn't seen yet") pulled 41%. Same email body. Same list. Different hook.
Was the AI subject line better? Arguably. Did a human have to choose it from a list of 20 options, tweak the phrasing, and decide it fit the audience? Absolutely. The AI generated the raw material. The human made the decision.
Personalisation at scale. This is where the maths gets interesting. Personalising an email beyond the first name, incorporating industry-specific language, company size references, local geography, and relevant pain points, takes a copywriter about 15 minutes per variation. When you're sending to 500 contacts across 8 segments, that's 120 hours of writing. Nobody has that time.
AI can produce those segment-specific variations in minutes. Not perfect ones. First drafts that capture roughly the right tone for each segment and hit the key personalisation points. A human then spends maybe 2 minutes per variation tidying up, adjusting tone, and catching the inevitable odd phrasing. Total time: a fraction of the manual approach, with a result that's arguably more consistent.
A/B testing at volume. Traditional A/B testing means writing two versions of an email and seeing which performs better. AI lets you test five or ten versions without the time investment of writing each from scratch. More variants means faster learning about what resonates with your audience.
Now the uncomfortable part. Because if you've been anywhere near LinkedIn this year, you've seen the results of people using AI without understanding its limitations.
Tone and authenticity. AI writes in a voice that sounds like a confident middle manager giving a TED Talk. It's fluent, it's structured, and it's completely devoid of personality. Every sentence is equally weighted. Every paragraph feels the same. There's no rhythm, no edge, no humanity.
This matters enormously in B2B email. Your recipients are humans who receive 50+ emails a day. They can smell corporate template copy from the subject line. If your email reads like it was generated by a machine, it'll be treated like it was generated by a machine: deleted without reading.
I made this mistake myself early on. I let an AI tool draft a full email for a client campaign, did minimal editing, and sent it. The open rate was fine (subject line carried it), but the reply rate was close to zero. When I actually read the email back as a recipient would, I understood why. It was perfectly competent and perfectly forgettable. No voice. No opinion. No reason to reply.
Understanding your specific market. AI knows about markets in general. It doesn't know about your market specifically. It doesn't know that your prospect's industry just went through a regulatory change, or that their main competitor launched a new product last month, or that the person you're emailing has been vocal on LinkedIn about cost pressures.
That contextual knowledge is what makes an email feel relevant. And it's the one thing AI consistently misses. It'll write a grammatically perfect email that says nothing specific enough to make the reader feel understood.
Knowing when to shut up. AI doesn't self-edit. It doesn't know that your email is too long, that the third paragraph repeats the first, or that the call to action is buried under four sentences of filler. It'll write as much as you let it, and every word will sound reasonable. But reasonable isn't the same as effective.
The best emails I've seen are short, specific, and leave space for the reader to respond. AI naturally writes long, general, and comprehensive. Those instincts are opposed.
Here's the bit that most AI evangelists gloss over: the editing.
Raw AI output needs heavy revision. I don't mean fixing typos. I mean restructuring paragraphs, cutting 40% of the content, injecting personality, adding specific references, and making the whole thing sound like it was written by someone who actually gives a damn. On a bad day, editing AI output takes longer than writing from scratch. On a good day, it saves maybe 30% of the time.
The people getting the best results from AI in email marketing aren't the ones pressing "generate" and sending. They're the ones using AI as a starting point, a creative partner that handles the blank-page problem and produces raw material that a skilled human then shapes into something worth reading.
AI in email marketing is improving fast. Six months from now, the tone problem will be less severe. The tools are already getting better at mimicking specific voices when given enough examples. Personalisation will get more sophisticated as AI models learn to pull in real-time data about companies and individuals.
But the fundamental dynamic won't change. AI handles scale. Humans handle substance. The combination works. Either one alone doesn't.
If you're not using AI in your email marketing at all, you're probably spending too much time on tasks a machine could handle. If you're using AI without significant human editing, you're probably sending emails that nobody reads twice.
The sweet spot is in between. Use the machine for what it's good at. Do the thinking yourself.
For full transparency, here's roughly how I use AI in email campaigns today.
I write the strategic brief myself: who we're targeting, what we want them to do, what the key message is, what tone fits this audience. That's the human thinking that AI can't replicate.
I use AI to generate first-draft copy, subject line options, and segment-specific variations. This gets me 60% of the way there in about 10% of the time.
I then edit heavily. Every email gets a human pass for tone, specificity, length, and authenticity. I add the market-specific references, the personality, the opinion that makes it sound like a real person wrote it.
Finally, I use AI again for testing: generating additional subject line variants, suggesting send-time optimisations, and analysing early results to inform the next send.
It's not glamorous. It's not "AI writes everything and I go for lunch." But it works. The output is better than pure human effort and dramatically better than pure AI output. That's the honest answer most people don't want to hear.
Martin Dugan, AA2
