The automation mindset: which tasks are actually worth it

The fastest way to waste a month is to automate the wrong task really well. I have done it, so now I run every idea through the same short filter before writing a line of code.

The three questions

Before automating anything, I ask:

1. 01How often does it actually happen? Not how annoying it is, how frequent. A painful task you do twice a year is a bad target.
2. 02Is the process stable? If the steps change every week, you will spend more time maintaining the automation than you saved.
3. 03What happens when it fails? Automation fails silently and at scale. If a failure is expensive and hard to notice, you need guardrails before you need speed.

If a task is frequent, stable, and safe to get wrong occasionally, it is a strong candidate. Miss any one of those and I think twice.

Do not automate a mess

Here is the trap. When a workflow is messy, automation feels like the fix. It is not. You are just making the mess run faster.

Fix the process on paper first. If you cannot explain it in five plain steps, it is not ready to be automated.

Clean, then automate. Every time I skipped the clean step, I regretted it.

Start with the boring 80 percent

The highest return is almost never the clever, edge-case-heavy part. It is the boring, high-volume middle. Copying data between two tools. Sending the same three follow-ups. Formatting a report the same way every Monday.

Automate that first, watch it for a week, and only then reach for the hard parts. The unglamorous wins compound quietly, and they rarely break.

A quick gut check

If you can describe the task to a new teammate in two sentences and they could do it correctly, a machine probably can too. If it takes you ten minutes and a lot of "it depends," leave it with a human for now.