The quiet cost of automating the wrong thing

When an automation goes wrong in an obvious way, you find out quickly. A refund fires twice. A confirmation email goes to the wrong list. Something visible breaks and someone shouts about it and you fix it. That kind of failure is annoying but it's honest. You know where you stand.

The failures that hurt businesses aren't like that. The ones that hurt are the automations that quietly succeed at the wrong job.

A team I worked with recently had automated their customer follow-up. Every closed ticket got a well-crafted, on-brand email a week later checking in. On paper it was a win. Consistent, on time, zero effort. Their support lead was proud of it and I could see why.

Then we looked at what happened after those emails went out. Nothing. No replies. No conversations. No new work coming back. Meanwhile the metric they actually cared about, repeat business from existing customers, had drifted down over eighteen months and nobody could quite explain why.

Here's what had happened. Before the automation, that follow-up was one of two or three people picking up the phone when they had a spare half hour. They didn't do it consistently. They did it selectively, to the customers they remembered, in a tone that sounded like them, and they got into actual conversations. Those conversations turned into referrals and second projects.

The automation was more consistent than the humans had ever been. It was also less useful, because the value was never in the touchpoint. The value was in the conversation, and conversations don't start from templated emails at scale.

This is the pattern I keep seeing. A business identifies something inconsistent and manual, replaces it with something consistent and automated, and celebrates the efficiency win. Six months later the number they were actually trying to move has quietly gone the other way, and nobody connects the two because the automation is still doing its job. It's doing exactly what it was told to do. It just wasn't the right job.

The uncomfortable truth is that a lot of the inconsistent human behaviour inside a service business is doing more work than it looks like it is. When your best account manager sends a personal note every Christmas but the others don't, that inconsistency is not a bug you should fix by automating a note from all of them. The note wasn't the point. The person sending it was the point, and everyone on the receiving end knew the difference.

So before you automate a touchpoint, ask a slightly different question than the usual one. Not 'how can we do this consistently at scale', but 'what is this touchpoint actually doing for us, and would a consistent version of it still do that?'. Sometimes yes. Sometimes the whole reason it worked was that a specific person chose to bother.

Automations are cheap to build now, so the discipline shifts. The valuable skill isn't building them. It's knowing which ones to leave alone. That's harder than it sounds and it's where most of the money is.

If you're staring at a list of things you could automate and aren't sure which ones will quietly cost you more than they save, that conversation is exactly the kind we like to have. Come and find us.