Before you automate it, question it.

Automation is easy to describe as a time-saving exercise. The better question is whether the task should exist in its current form at all.

Follow one real piece of work

Pick an enquiry, invoice or report and follow it from start to finish. Write down who touches it, what they change, where they wait and what they copy. Ask people to show you the process rather than describe the official version.

The gaps are usually revealing. A field is typed twice because two tools do not connect. An approval waits because ownership is unclear. A report is assembled every week because nobody has asked whether anyone still reads it.

Separate rules from judgement

Predictable rules make good automation candidates. If an approved record contains these fields, create that task. If a date passes, send a reminder. If a number does not reconcile, flag it for review.

Judgement is different. Deciding whether an unusual customer request is reasonable involves context. You can automate the gathering of that context without pretending the decision is mechanical.

Choose the smallest useful change

Make failure visible

An automated task that silently fails is often worse than a manual task that someone notices is missing. Define the error path alongside the successful path: who sees the failure, what information they receive and how they can recover.

Before changing the process, record a baseline. How long does the task take? How often is it repeated? How much work returns for correction? Measure the same things afterwards. The useful result is a better process, not a larger collection of automations.