Automation conversations tend to start at the exciting end — dashboards, predictions, intelligent routing. The projects that actually pay for themselves usually start somewhere far duller.
Find the tasks nobody mentions
Ask a team what takes their time and they will describe the difficult parts of their job. The automatable work is the stuff they do not think to mention, because it feels like background noise:
- Copying figures from one system into a spreadsheet
- Compiling the same weekly report
- Chasing three people for an approval
- Sending the same reminder to customers with overdue balances
- Re-keying supplier invoices
None of it is interesting. All of it is measurable, rule-based, and repeated — which is exactly what makes it automatable.
Measure before you build
Before writing anything, establish how long the task takes and how often it runs. A task taking two hours weekly costs over a hundred hours a year. That number decides whether automating it is worth doing, and it is the only honest way to judge the result afterwards.
If nobody can estimate the time, that is itself a finding.
Automate the rule, not the exception
Most manual processes have a clean path covering the large majority of cases, plus exceptions that genuinely need judgement. Trying to automate the exceptions is where these projects overrun.
Automate the clean path. Route anything unusual to a person, clearly flagged. Handling the bulk of cases reliably beats handling all of them unreliably.
Build in visibility from the start
Silent automation is dangerous. When a scheduled job fails at 3am, someone needs to know before a customer does. Every automated process should record what it did and alert a named person when it could not.
An automation nobody is monitoring will eventually fail quietly, and the damage compounds until someone notices by accident.
Expect the process to change
Automating a process usually reveals that parts of it existed only to compensate for the manual method. A reconciliation step may become unnecessary once transfers stop being manual.
Be prepared to simplify the process rather than faithfully reproduce it. Automating an unnecessary step just makes it happen faster.
A realistic first project
Pick one task. Confirm its cost. Automate the clean path. Add monitoring. Measure again after a month.
That cycle is unglamorous and it works. It also funds the next one from real savings rather than a projection.
Our business automation work starts with a short process audit for exactly this reason — the highest-value target is rarely the one people expect.
