Automation without authenticity is just spam with better graphics

Every serious email tool now sells automation as a headline feature. Behavioural triggers. Predictive send times. AI-generated subject lines. Personalised sequences that adapt to what the subscriber clicked on eighteen days ago. The pitch is that automation lets you scale — that the program can grow without the writing having to grow with it.

This is technically true and, in the way it is usually deployed, catastrophically wrong. Automation is a fine tool for the parts of an email program that are legitimately mechanical. It is a terrible tool for the parts that are relational. Getting the line between the two right is one of the most important editorial decisions a program makes, and one that almost no team makes deliberately.

What automation is good at

Automation is good at things that are the same every time, that are triggered by observable actions, and that the subscriber understands are automated. A password reset. An order confirmation. A shipping notification. An abandoned cart reminder. A subscription renewal notice. These messages should be automated. Automating them is not a compromise. It is the correct architecture.

The subscriber, in these cases, expects the message and understands its automation. Nobody feels manipulated when Amazon sends them a shipping notification. The message is doing exactly what the subscriber expects it to do, in the way they expect it to do it. Automation here is transparency, not sleight of hand.

What automation is bad at

Automation is bad at things that are supposed to feel personal, that are triggered by inference rather than direct action, and that the subscriber does not know are automated. The birthday email. The "we noticed you were reading about X" email. The "your customer success representative wanted to check in" email that is signed by a person who has never touched the account.

These messages are the ones that produce the specific, sinking feeling every subscriber knows — the moment you realise a message that seemed personal was actually a template. That feeling costs trust. It costs trust cheaply if it happens once. It costs trust expensively if it becomes a pattern. And the tools that make automation easy also make this pattern easy to fall into without noticing.

The subscriber can tell

One of the most important shifts of the last several years is that subscribers have become expert readers of automation. Everyone who has an inbox has received thousands of automated emails. Everyone has learned to spot the specific tells — the merge-tag greeting that reads as merge-tag, the "quick question" subject line that has been sent to ten thousand people, the reply-to address that is a role account rather than a person.

The assumption that subscribers cannot tell is out of date. They can. The programs that succeed now are the ones that either automate transparently — the subscriber knows this is a shipping notification and it is fine — or write personally, at whatever scale their team can sustain.

The middle path

There is a middle path, and it is where most well-run programs actually live. Automate the mechanical. Hand-write the relational. When these two are clearly separated, both work. When they are muddled — automation that pretends to be personal, personal touches that get folded into an automated sequence — both fail.

Practically, this means the welcome email should probably be hand-written and updated when it starts to feel dated. The password reset should be automated. The Sunday newsletter should be hand-written each week. The order confirmation should be automated. The "we noticed you have not opened in a while" re-engagement should be hand-written, or at least written recently by a person, even if it goes out through an automated trigger.

Automate everything the subscriber knows is automated. Hand-write everything else. Any other split is asking the subscriber to notice the seam.

What to audit for

The test I run against my own program every few months is straightforward. If a subscriber replied to any given email, would a real person answer? For the automated ones — the shipping notifications and the password resets — the answer can be "the reply goes nowhere and that is fine because the message is clearly transactional." For the personal ones — the newsletter, the welcome, the re-engagement — a real reply had better reach a real person.

If your program cannot answer this cleanly for every email in the sequence, some of the email is falling into the middle zone that reads as fake. Cleaning up that middle zone is the highest-leverage editorial move most programs can make. It usually takes one afternoon and pays back for years.

The tools will keep getting better at making automation invisible. The subscribers will keep getting better at spotting it. The gap between those two curves is where the future of email marketing gets decided, and I think the writers who lean toward the hand-written side of the line will win the next decade.