What to automate firstWhat to automate first

A process is a good automation candidate when it meets three conditions at once: it repeats often, it follows clear rules, and today a person does it by copying information from one place to another. Miss any of the three and automating it will probably cost more than it saves.

The temptation is to start with the most visible or the most irritating process. Neither tends to be the most profitable one.

For each process, multiply three things: how many times a week it happens, how many minutes it takes each time, and what happens when someone gets it wrong. That third factor is the one most people skip, and usually the one that weighs most.

A two-minute task done forty times a week that under-invoices when done wrong is worth far more than a one-hour task that happens monthly.

In most of the companies we work with, the first four candidates are the same ones:

A process that changes every month, depends on the judgement of an experienced person, or that nobody can explain end to end, is not ready. Automating a process you do not understand only makes the errors happen faster and in greater volume.

In those cases the first step is not technical: it is writing the process down and stabilising it. Often that exercise alone solves half the problem, with no software involved.

The first automation should be running in weeks, not months, and its effect should be visible without a dashboard: somebody stops doing something. That concrete result is what funds and orders everything that comes after.

Write down how long it used to take. It is the only way to defend the next investment with a number instead of an impression.