How do you decide what to automate first?

Most automation projects fail at the moment they are chosen, not the moment they are built.

People automate what annoys them most. Annoyance and cost are different things. Here is the test, and it takes five minutes.

1. How often?

Daily is gold. Weekly is worth a look. Monthly rarely pays back unless it takes a full day. Twenty minutes a day is eighty hours a year. A day each quarter is not.

2. Judgement or attention?

If a competent new starter could do it correctly from a checklist, strong candidate. If doing it well needs ten years of knowing which customers are difficult, be careful — you are building a tool for a human, not a replacement.

3. What does failure cost?

A process that fails invisibly is more dangerous automated than manual. A person notices something is off. A script does not, unless you make it. If failure is expensive, budget for the checking as well as the doing.

4. Does anyone use the output?

This is the one that catches people. I have watched a business plan to speed up a report that, on investigation, nobody read. The fastest process is the one you delete.

The trap

Automating a bad process instead of fixing it. If someone re-types an order because two systems do not talk, the answer might be a two-hour integration, not a clever workaround that makes re-typing permanent.

Always ask why the step exists before making it efficient. Sometimes the win is subtraction, and subtraction is free.