Counting the real cost

The arithmetic everyone does

“This takes me ten minutes a day. That is fifty minutes a week, forty hours a year. I should automate it.”

That arithmetic is almost always wrong, in both directions, and getting it right is what separates automation that pays from automation that quietly costs.

What the ten minutes leaves out

Switching cost. A ten-minute task that interrupts focused work costs far more than ten minutes. If it lands mid-morning and pulls you out of something hard, the real cost may be forty minutes. This makes the case for automating stronger than the raw number.

Frequency versus stakes. A task you do once a month for two hours is eighty minutes a month of work. A task you do forty times a day for fifteen seconds is ten minutes a day, but it also generates forty chances to make a mistake. Count errors, not just minutes.

What automation costs that people forget

Building it. Usually estimated correctly and usually the smallest number in this list.

Maintaining it. Every automation is a small piece of software that now has to survive password changes, upgrades, moved files, changed formats, and a network that goes down. Budget real hours a year, forever.

Watching it. An automation you do not monitor is not saving you work; it is deferring it. When it fails silently for six weeks, you pay all six weeks at once, usually at the worst possible moment.

Losing the skill. Two years after you automate something, nobody remembers how to do it by hand. When the automation breaks, the fallback is not “do it manually” — it is “nobody here knows how this works.” This cost is invisible until the day it is enormous.

An honest threshold

My rough rule, and it is rough: automate when the task is frequent, stable, and consequential — you do it often, the steps have not changed in months, and getting it wrong actually matters. Two out of three is a maybe. One out of three is a no.

“Consequential” is the one people drop, and it cuts both ways. A frequent, stable task that does not matter is not worth the maintenance. A rare task that matters enormously — a restore from backup, say — may be worth automating precisely because you do it so rarely that you would otherwise fumble it under pressure.