Cut Before You Automate

01.Everyone Wants to Automate

Everybody wants to automate. Almost nobody wants to subtract. So they end up with a beautifully automated mess — a machine that does the wrong things very efficiently, forever.

I get the appeal. Automation feels like progress. You connect two tools, watch a task do itself, and get a small hit of "I'm running a real business now." It's satisfying. It's also, most of the time, the second thing you should be doing — and people keep reaching for it first.

02.What automation actually does

Automation is an amplifier. That's the whole truth of it. It takes whatever process you point it at and makes it run faster, more often, with less friction. Point it at something clean and it's a superpower. Point it at something broken and you've simply built a faster, more reliable way to lose.

Worse: once a bad process is automated, it becomes invisible. It hums away in the background, doing its wrong thing on schedule, and you stop questioning it — because now it "just works." Automation doesn't fix a flawed step. It preserves it and hides the evidence.

03.The cowpath problem

There's an old planning story about a town that paved its footpaths by simply laying concrete over the muddy tracks people had already worn across the fields. The paths were convenient once. They made no sense as permanent roads — but there they were, set in stone, curving around fields that no longer existed.

That's what automating too early does. You take the messy manual route you happened to fall into — six steps, three tools, a bit of copy-paste — and you pave it. Now it's permanent. And the four unnecessary steps are just as permanent as the two that mattered.

04.What most people do

They map their current process, wire every step of it into an automation, and feel enormously clever. A member of ours did exactly this last year — built a slick six-step onboarding flow, fully automated, genuinely impressive to look at.

When we sat down and actually walked through it, four of the six steps existed only because the fifth one was in the wrong place. Move that one step, and the other three vanished. The winning move wasn't a better automation. It was a pair of scissors.

05.Stack Hygiene — the three cuts

Before you automate anything, run it through what I call Stack Hygiene — three cuts, in this order:

Delete. Does this step need to exist at all? Not "can it be faster" — does it need to happen? Be ruthless. This is where most of the savings live, and it's the cut people skip because deleting doesn't feel like building.

Merge. Are two tools doing one job? Are two steps really one? Every tool you cut is a subscription, a login, a point of failure and a small monthly phantom load, gone.

Simplify. Of what survives, can the path be shorter? Fewer handoffs, fewer fields, fewer decisions?

Only now — on what's left after those three cuts — do you automate. And you'll find there's far less of it, it's far cleaner, and it actually deserves to be made permanent.

Not sure which parts of your stack to cut and which to keep? The Backend GPS scorecard maps your tools and workflows and shows you where the redundancy — and the leaks — actually are, before you spend a penny automating.

06.The order that matters

Eliminate, then simplify, then automate. In that order, every time. Reverse it and you'll spend money and hours making your worst habits permanent.

The point of a good backend was never "more machines." It was less to think about. Automation is one tool for that — the last one you reach for, not the first. Subtract until what's left is clean. Then let a robot run it while you go and do the work only you can do.