The short version: Automate data entry, segmentation, send times, and subject-line testing with AI, but never automate the opening line, the final offer, or the decision to send to someone at all. Your email list is your most valuable asset and one bad AI decision can crater it in hours.
I made a horrible mistake in 2019 that cost me roughly 12,000 pounds in lost revenue. I handed AI the keys to my welcome sequence and told it to "optimise engagement." Within four days, the system had automatically removed 4,200 people from my list because their first click came slower than some arbitrary threshold. I didn't notice for a week. By the time I saw the damage, those people were gone, and I had no way to win them back.
That experience is why I'm blunt about this: email is too important to autopilot fully. But there are things AI is better at than you are. The trick is knowing where to draw the line.
What AI Should Automate in Email Right Now
1. Data entry and list hygiene
If you're manually checking for bounced addresses, duplicate emails, or formatting inconsistencies, AI is faster and more thorough than you'll ever be. Tools built on AI can flag addresses that look fake, catch cases where someone signed up as "XxSuperCool99xX" (probably a test), and merge duplicate contacts automatically. This saves you hours and keeps your sender reputation clean. A bad sender reputation means your emails go to spam; a good one means deliverability above 95 per cent.
Step one: Run your list through an AI-powered validation check monthly. Step two: Set rules so that addresses with zero opens after 12 months get a re-engagement campaign automatically. Step three: Only remove them from the list if they don't open that either.
2. Segmentation and dynamic list building
AI can watch your subscriber behaviour in real time and sort people into buckets based on what they do, not what you think they do. Someone who clicks every link about video content goes into the video bucket. Someone who only opens on Tuesdays gets Tuesday sends. Someone who bought your course three months ago but never logged in gets a "let's help you start" message.
This is where AI excels because humans miss patterns. You could spend three hours manually building segments; AI does it in seconds and gets it right. The data I see from my own lists shows that segmented campaigns perform 18-24 per cent better than blasts, and AI speeds that up.
3. Send-time optimisation
AI knows that you, a UK-based coach, probably shouldn't send at 6am to your US list. It also knows that your best opener might be 9am for one person and 2pm for another. Let AI pick the send time. It's a clear, data-driven call with no emotional stake. Send at the time your person is most likely to open.
4. Subject-line and preview-text testing
This is one of the few places where I trust AI fully. AI can generate 15 subject-line variations, test them on a small sample of your list, measure opens, and pick the winner automatically. It removes the human bias that makes you think a clever pun will work when it won't. I've tested this hundreds of times and AI beats my own instinct about 60 per cent of the time.
Real example: Last year, I was launching a workshop. My instinct said the subject line should be "The Three-Day Email Intensive: Limited Seats." AI suggested "I'd book this if I were you." The AI version won by 23 percentage points. I would have left 23 per cent of opens on the table.
5. Cart-abandonment and behaviour-triggered sequences
If someone adds something to their cart and walks away, AI should send them a reminder after 4 hours, then 24 hours, then 72 hours. The message can be templated and AI-written because the rules are fixed and the situation is clear. Same for "you watched the webinar but didn't buy" or "you engaged with this resource but never clicked the next one."
The key is that the trigger is mechanical (they did X action) and the response is templated (send message Y). AI doesn't need to think here. It just needs to follow a rule.
What You Must Keep Human
The first line
Your opening line is the relationship. It's where you show that you know something about the person on the other end. "Sarah, I saw you watched the entire free masterclass" is human. "Dear valued customer" is what AI does when it's pretending to be human and failing. Never automate this. Read your analytics, know your audience, and write it yourself or hire someone who will.
The actual offer or ask
This is the moment where money either changes hands or your relationship deepens. If you let AI write your core offer without human oversight, you will sound generic. I've seen AI-generated email copy that was technically correct but so bland that I couldn't imagine anyone clicking through. An offer needs conviction, specificity, and your voice. "Buy this course" doesn't work. "I spent six months on this and it works because I tested it on 400 people" works.
The decision to send at all
This is the honest point that most articles don't make: some emails should never be sent. A good AI system will flag when you're about to email a cold list, or when you're sending the fifth message in a week, or when you're re-mailing someone who unsubscribed last month. But the final decision to hit send must be human. I've been about to send something that would have damaged my reputation, and only because I paused did I catch it.
Rule: Always ask "Would I want to receive this email?" before you send. If the answer is no, don't send it. AI won't save you from that mistake.
Anything that touches your brand voice
Your emails are one of the last places where customers see the real you. If AI writes them all, you become one of 500 similar voices. Automate the data work. Write the connection work yourself. Think of it this way: you wouldn't let AI pick your friends or your spouse. Don't let it write the words that build your business relationships.
The Real Workflow: What Hybrid Looks Like
Here's how I run my email now, and it's the template I recommend:
- AI validates and segments the list automatically (15 minutes of setup, then hands-off).
- AI tests subject lines and preview text on a sample (5 per cent of the list, measured after 12 hours).
- AI identifies the best send time for each person and schedules the send window (no human input needed).
- I write the opening line and the core offer in my voice (45 minutes).
- I review the email one final time before any send goes out (5 minutes).
- AI monitors opens, clicks, and bounces and flags anything unusual (ongoing, alarms only).
This takes me about an hour per campaign instead of three, and the results are better because I kept the thinking parts and handed off the grunt work.
Why This Matters More Than You Think
Your email list is not a marketing channel; it's a permission asset. People gave you permission to show up in their inbox because they believed something good would come from it. The moment you violate that with over-automation, bad copy, or AI-generated mediocrity, they're gone. And unlike social media, you can't get them back. They won't follow you again.
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.
That's why I don't take this lightly, and why I'm direct: use AI to handle the mechanical work so you have time to handle the human work. But don't outsource the relationship. That's where the money lives.
If you want a broader look at how AI is reshaping marketing automation, I've written about AI automation for marketing teams and real examples that save time. And if you're thinking about building more sophisticated systems, a guide to building your first AI agent for marketing might help you think bigger without losing the human edge.
I have written more around this on the site: What Works: Picking a Social Media Marketing Tool That Won't Waste Your Money, The AI for Marketing Era Is Quietly Ending. AI Agents Are Taking Over, AI Automation for Marketing Teams: Real Examples That Save Hours Every Week.
The Subject Line Test I Run Before Any AI Copy Goes Out
Here is something I tell every client who wants to hand subject lines over to AI completely: I still run a 24 hour split test on anything the tool generates, even after three years of using these platforms. Last quarter I let an AI tool write subject lines for a webinar reminder sequence. It produced "Don't Miss Out: Your Spot Is Waiting" for one send. My own version, written in ten minutes, was "I saved you a seat, quick question though." Mine pulled a 41% open rate against the AI version's 27% on a list of just under 9,000 subscribers. The AI line was not wrong, it just sounded like every other AI line, and my subscribers have gotten good at spotting that tone.
What I have learned from tracking this across dozens of campaigns is that AI-generated subject lines tend to cluster around the same three or four patterns: the urgency play, the curiosity gap with a colon, the "here's why" structure, and the numbered list tease. Once you have seen a few hundred of them, you start recognizing the rhythm, and I suspect a good chunk of your audience does too even if they could not name why a subject line feels off. The fix is not to abandon AI for this task, it is to use it for volume and then inject one specifically human sentence, usually something with a small personal detail or an odd verb, before you send.
My actual workflow now: I ask the AI for ten subject line variations, I throw away the top three because they are always the most generic, and I rewrite whichever one is left standing so it includes a specific number, a name, or a slightly imperfect phrase. Then I test that against one fully AI version on a 10% sample split before committing the winner to the full list. This adds maybe fifteen minutes per campaign. Over a year of weekly sends, that is roughly thirteen hours of extra work, and in my case it has consistently added three to six points of open rate compared to sending the AI's first choice untouched.
If you only take one thing from this, take the sample split test. Most people either automate subject lines fully or skip AI for them entirely. The middle path, where AI does the brainstorming and a human does the final edit and a small live test decides the winner, is the one nobody talks about because it is not as clean a story as "just automate it" or "never automate it."
Frequently Asked Questions
Can AI write the entire body of an email if it's just informational (not a sales email)?
Rarely should it. Even informational emails benefit from your voice and specific examples. If you're sharing a tip, a story, or a resource, that's still an opportunity to build trust. A templated, AI-written info email feels like a notification, not a message from someone who cares. Where I do use full AI generation: transactional emails (order confirmation, password reset) because those don't need personality. Everything else should have your hand in it.
What's the biggest risk of over-automating email with AI?
List decay. You send more emails because AI makes it easy, recipients stop engaging because the content feels generic, your open rates drop, email clients flag you as spam, and suddenly 20 per cent of your emails aren't reaching the inbox at all. I've seen this happen to businesses that went all-in on AI email with no human review. They gained efficiency and lost revenue.
Should I use AI to decide who to remove from my email list?
No. AI can flag who looks inactive (no open for 180 days), but the removal decision is yours. Send them one re-engagement email written by you, asking if they still want to hear from you. If no response, remove them. If AI decides without asking, you might accidentally delete someone who was just on holiday or between jobs.
Is there any part of email marketing where AI should be 100 per cent in charge?
Yes: list validation and segmentation. Let AI own that completely. Bad data will tank everything downstream, and AI is better and faster at catching it. For everything touching tone, relationship, or the actual message, keep human judgment in the loop.
Related reading: Lead Gen Content Marketing: How to Stop Creating Content Nobody Acts On and How to Earn Passive Income: What Works (And What's Bollocks).
Want the complete version? Read where I break down AI marketing.
