What should you automate first?

Start with what repeats

A task you do once a quarter is not worth wiring up, however tedious it is. A task you do eight times a day, the same way each time, is. Count how often it actually happens over a normal week before you decide it is a problem worth solving.

Then check the rules are clear

If you can write down what to do in a handful of if-this-then-that lines, a machine can follow it. If the answer is genuinely "it depends on the customer", that judgment is the part you keep. Automate the fetching, sorting, and drafting around it instead.

Then decide where a person checks the output

The steps worth automating first are the ones where a person still sees the result before it leaves the building. That is not a compromise. It is what makes the thing safe to turn on while you still trust it, and it is the difference between a workflow you keep and one you quietly switch off after a bad week.

Why one workflow, not five

Picking one process and finishing it teaches you more than starting five. You find out where your information actually lives, which tool refuses to talk to the others, and how much of the work was really judgment. That knowledge makes the second workflow much faster.

Written by Joel Moreno of Ask Mr. JM. The presenter image used across this site is an AI likeness; the advice is his own.  All lessons