When is a process worth automating

Not every manual task is worth automating. Some are rare, some change every month, some take a minute and will never take more. Automating those costs more than it saves. Before building anything, it is worth running the task through a short test.

The four things that decide it

Frequency. How often does it run — daily, weekly, every month-end? The more often, the more a one-time build pays back. A task done once a quarter rarely justifies the effort.

Time per run. How long does it take a person each time, start to finish, including the interruptions it causes? Multiply by frequency. Two hours every week is a hundred hours a year — a week and a half of someone’s time on one task.

Cost of errors. What happens when it goes wrong? Re-keyed figures that feed a tax return or a board report carry a cost far beyond the minutes to fix the typo. High error-cost moves a task up the list even if it is quick.

Stability. Does the task work the same way each time, or does it change constantly? Stable, rule-shaped work automates well. Work that depends on fresh judgment every time does not — and shouldn’t.

A task that scores high on the first three and is stable is a strong candidate. A task that is frequent and slow but changes every run is usually better fixed by changing the process first, then automating.

A quick way to estimate payback

Take the hours the task costs per year (time per run × runs per year). Compare that to the one-time cost of building the automation plus its running cost. If the build pays for itself inside a year, it is almost always worth doing; the saving compounds every year after. If payback is two years or more, look harder — there may be a smaller piece worth automating instead of the whole thing.

When the answer is no

• It runs rarely and will keep running rarely.
• It changes so often that the automation would need rebuilding each time.
• It is really a decision, not a task — the “work” is the judgment, and that belongs with a person.
• The process itself is broken. Automating a bad process just makes the mess faster. Fix it first.

Being told honestly that a task is not worth automating is worth as much as a build. On a scoping call we will tell you which of your tasks clear this bar and which don’t — including the ones we would leave alone.