What to automate first: a simple way to choose

The short answer

Pick the process that scores highest on three things: how often it is repeated, how clear its rules are, and how much a delay costs you. Leave the rare, the fuzzy and the harmless for later.

Most companies start in the wrong place. They automate what is interesting, or what the newest tool makes easy, instead of what the business actually waits on every day.

List the work, not the tools

Write down what your team does repeatedly in a normal week: answering the same enquiries, copying orders into another system, sending quotations, reminding customers, preparing the same report. Do not mention software yet.

Next to each item write three things: roughly how many times a week it happens, whether a new employee could do it from a one-page instruction, and what goes wrong when it is done late.

The three tests

Frequency: something done many times a day pays back sooner than something done once a month, however painful the monthly task feels.

Clarity: if two of your own people would handle the same case differently, the process is not ready to automate. Agree the rule first. Automation makes a clear process fast and a confused process confusing at speed.

Cost of delay: an enquiry answered tomorrow may be a customer lost. A report delivered tomorrow is usually just a report delivered tomorrow.

Finish one before starting two

Automate one process from its first step to its last, including what happens when something fails. A workflow that works on a good day and stops silently on a bad one is worse than doing the work by hand, because nobody is watching it any more.

When the first process has run for a few weeks without surprises, the second one is easier: the connections to your systems already exist and your team has seen what to expect.