A few months ago, a client replied to one of my onboarding emails saying it was "the warmest, most personal welcome she'd ever gotten from a business." I hadn't written a single word of it. AI had — using a system I'd spent about two hours setting up. That's the promise of AI email automation when it's done right. The problem is, most people do it very wrong.
I've seen entrepreneurs automate their customer emails and watch open rates tank, replies dry up, and unsubscribes spike — not because automation is bad, but because they skipped the step that makes the whole thing work. This post breaks down exactly what I do, what that critical step is, and how you can have a genuinely human-feeling email system running by next week.
The mistake everyone makes first
Most people approach AI email automation like this: they open ChatGPT or Claude, type "write a welcome email for new customers," copy whatever comes out, and call it done. Then they wonder why it feels flat. The output is only as good as the input you give it. A generic prompt produces a generic email — and generic emails get ignored.
The fix isn't more AI. It's a voice document. Before you write a single prompt or set up a single automation, you need to spend 20 minutes writing down how you actually talk to customers. Include phrases you use regularly, things you'd never say, your typical tone (formal? casual? warm but direct?), and a few examples of emails you've sent that you were happy with. Paste all of that into your AI tool and tell it: "This is my voice. Every email you write for me should sound like this." That one step is what separates AI emails that feel robotic from ones that feel like you.
The voice document is your secret weapon. Keep it in a running doc and update it as your communication style evolves. Every AI session starts with pasting it in — this 30-second habit is what makes everything else work.
The five emails worth automating first
Not every email is worth automating. I'd focus your energy on the ones that repeat constantly and carry the most weight in the customer relationship. In my experience, five types pay off immediately.
The welcome email is the highest-leverage place to start. It goes out right after someone buys or signs up, when their attention and goodwill are at their peak. A warm, personal welcome sets the tone for everything that follows. After this, the onboarding sequence — two or three emails that walk new customers through getting value fast — can basically run itself once you've written the logic once.
The check-in at the 30-day mark is one most businesses skip entirely, which is exactly why doing it makes you stand out. A short, genuine "how's it going?" email with a specific question gets replies that give you customer intelligence you'd otherwise have to pay for. Then there's the re-engagement email for customers who've gone quiet, and the review or referral ask that most people send too late (or never send at all). These five, automated and running in the background, cover the entire customer lifecycle.
A prompt that actually works
I want to show you what a good AI email prompt looks like versus a bad one, because the difference is significant. A bad prompt is: "Write a welcome email for a new customer." A good prompt looks more like this:
Example promptThat level of specificity is the difference between an email your customers save and one they delete. I've used variations of that structure for every email type I mentioned above — the logic stays the same, you just swap in the context.
Connecting it to a real workflow
Writing good AI email drafts is only half the job. You also need a way to trigger the right email at the right moment without manually doing anything. The simplest setup I've found uses three tools: your email platform (I use ConvertKit, but this works with Mailchimp, ActiveCampaign, and most others), a trigger event (purchase completed, form submitted, tag applied), and a pre-written email sequence loaded and ready to go.
Here's what my process looks like in practice. I use AI to draft all my email sequences once, in batches. I set aside one afternoon every quarter, paste in my voice document, and write or refresh every automated email I need. Then I load them into my email platform, map the triggers, and test everything by going through the purchase or signup process myself as a fake customer. Once it's working, it genuinely runs without me. New customer signs up on a Tuesday at 2am — they get my warm welcome email at 2am Tuesday. I'm asleep. It doesn't matter.
One thing I check quarterly: I forward myself three or four of the automated emails and read them fresh, imagining I'm a new customer receiving them. If anything sounds dated, off-tone, or generic, I spend an hour updating it. Fresh eyes every 90 days keeps the automation feeling alive.
The part AI can't do for you
I want to be honest about what automation doesn't replace. When a customer replies to one of my automated emails — and they do, regularly — that reply comes to me. Real conversation still requires a real person. The automation opens the door; you still have to walk through it when someone invites you in. The goal isn't to never talk to your customers. It's to make sure the right words reach them at the right moments, automatically, so that when they do reach out to you personally, it's because something real is happening.
I've also learned not to automate the truly sensitive moments. If a customer has a problem, a complaint, or is clearly frustrated, that email should come from you — or at least be reviewed by you before it goes out. AI is excellent at the warm, routine, positive touchpoints. It's less equipped to handle the moments where a customer needs to feel genuinely heard by a human being.
Where to start today
If you're starting from zero, here's the sequence I'd follow. Write your voice document — 20 minutes, no overthinking it. Then pick just one email to automate first: your welcome email. Use the prompt structure above, iterate on the draft until it sounds exactly like you, then load it into your email platform and connect the trigger. Live with it for two weeks. Watch what happens. Once you see it working, add the next email in the sequence. Build the system one piece at a time, and within a month you'll have something that makes your business feel more personal than most businesses that have real humans manually writing everything.
That's the counterintuitive truth about AI email automation done right. It doesn't make you sound less human. It makes you more consistently human — because the warm, thoughtful version of you shows up every time, not just when you happen to have the energy for it.