AI has become remarkably good at writing emails. Sometimes it's frighteningly good. Give it a product description, a campaign brief, and a target audience, and within seconds you'll have something that would have taken a copywriter an hour to produce.
For many emails, that's fantastic. Promotional campaigns, product summaries, welcome sequences, abandoned cart reminders, event notifications — these follow patterns, and patterns are exactly what language models are designed to recognise. But there's a growing temptation to let AI write everything. That's a mistake. Not because AI isn't capable. Because some emails aren't simply collections of words. They're moments that define how customers feel about your company.
And that still requires a human.
The Difference Between Information and Intent #
Two emails can contain identical information. One builds trust. The other destroys it.
The difference isn't grammar. It's intent. Humans communicate far more than words — confidence, uncertainty, accountability, empathy, excitement, vulnerability. AI imitates these emotions. Humans actually feel them. Sometimes that difference matters a lot.
Where Humans Still Need to Write #
The following emails require human authorship — or at minimum, heavy human editing. Not because AI can't produce them. Because the cost of getting them wrong is too high.
Apology emails and accountability messages. If your company has made a mistake, don't let AI apologise for it. A genuine apology should answer what happened, why it happened, who is taking responsibility, what changes next, and why customers should trust you again. Those answers require honesty, not prediction. The same applies to pricing changes, policy updates, and service disruptions — customers want to know who made the decision and why. AI cannot accept responsibility. Only people can.
We sincerely apologise for any inconvenience this may have caused.
You've read that sentence hundreds of times. It's technically correct. It's also emotionally empty. Nobody reads it and thinks, "They really mean it."
Founder letters. The best founder emails don't sound perfect. They sound human. Sometimes they're awkward. Sometimes they're emotional. Sometimes they admit uncertainty. That's exactly why people read them. Nobody subscribes for AI-generated authenticity. They subscribe for insight from the person building the company. Your perspective is the product. Don't outsource it.
Crisis communications. Security incidents. Service outages. Data breaches. These are not content marketing exercises. Customers aren't just looking for information — they're judging competence. Every sentence communicates something beyond its literal meaning. Too defensive, you look evasive. Too vague, you look unprepared. Too polished, you look insincere. AI can help structure these emails. It shouldn't own them.
Brand manifestos and company values. Every memorable company believes something. Patagonia. Basecamp. Notion. Their best writing doesn't just explain products. It explains philosophy. That philosophy comes from years of experience, decisions, and trade-offs. AI can summarise those beliefs. It shouldn't invent them. And ironically, company values are often the most AI-generated documents in existence — "Innovation. Integrity. Excellence." Almost every company says the same thing. Real values aren't slogans. They're difficult choices. The decision to delay a launch because quality wasn't good enough. The choice to lose a customer rather than compromise on ethics. Those are the stories worth telling.
Customer stories. AI can improve writing. It shouldn't manufacture experiences. Real customer stories contain details nobody would invent — small frustrations, unexpected successes, specific numbers, tiny moments. Those details create credibility. Removing them in favour of generic AI language usually makes the story worse.
Difficult pricing announcements. Price increases are never popular. The easiest email to generate is "To continue providing the highest quality service..." Customers have seen that template countless times. What they actually want is transparency. Explain why costs changed, what customers receive, when it happens, who it affects, and whether alternatives exist. People are remarkably understanding when they're treated like adults. AI tends to soften difficult conversations. Sometimes clarity is kinder.
Strategic decisions. Discontinuing a product. Entering a new market. Changing company direction. Acquiring another business. Readers aren't just evaluating the decision — they're evaluating leadership. That confidence has to come from somewhere real.
The "Why" Problem #
AI is exceptionally good at writing what. Humans still outperform it at explaining why.
Consider these:
AI: We're introducing a new dashboard to improve reporting.
Human: We realised customers were spending more time exporting data than analysing it. That felt backwards, so we rebuilt reporting from scratch.
Same feature. Different meaning. The second tells a story. The first announces functionality. When customers understand why something exists, they care about it more. That "why" comes from real experience — from knowing the frustration of watching users struggle with something you built.
The Trust Scale #
Think about your emails as existing on a spectrum. The more emotionally significant the message becomes, the less automation you should introduce.
| Type of Email | AI Ownership | | Weekly promotions | High | | Product updates | High | | Educational newsletters | Medium | | Product launches | Medium | | Customer success stories | Medium | | Founder updates | Low | | Pricing changes | Very Low | | Apology emails | Human only | | Crisis communications | Human only |
What AI Can Still Do #
This isn't an argument against AI. Quite the opposite. AI is brilliant at helping humans think — challenging assumptions, improving clarity, generating alternatives, fixing grammar, rewriting tone, and organising ideas. It's an outstanding editor. Editors have always existed. The author still matters.
The strongest email teams follow a simple process: AI handles research, first drafts, and option generation. Humans choose direction, edit heavily, own brand voice, write sensitive messaging, and make final decisions. The closer an email gets to trust, the more important the human becomes.
The Bottom Line #
AI should replace repetitive work. Not responsibility. Use it to eliminate blank pages. Use it to accelerate editing. Use it to explore ideas you might not have considered. But don't let it become your company's voice.
Customers don't build relationships with language models. They build relationships with people. The irony is that as AI-generated content becomes more common, genuinely human communication becomes more valuable. The future of email marketing probably isn't choosing between AI and humans. It's knowing exactly when each one should be writing.