Key takeaways
- Prioritise high-volume, repetitive, rule-based tasks — low-frequency work rarely justifies the build cost.
- Fix a process before automating it, or you will make the ambiguity faster rather than removing it.
- Design exception paths deliberately so unusual input surfaces to a person instead of vanishing into a log.
The instinct when a manual process feels tiring is to automate it. That instinct is right often enough to be dangerous, because the process itself may be the problem.
Volume Before Efficiency
A high-volume, repetitive, rule-based task is where automation reliably pays. A task performed twice a month rarely justifies the build cost, however tedious it feels.
Never Automate a Broken Process
If the process relies on undocumented knowledge — someone knowing which customer to escalate to, or which spreadsheet tab is authoritative — automating it makes the ambiguity faster and harder to remove. Fix the process first, then automate the corrected version.
Where the Gains Appear
- Data entry and transfer between systems that do not talk to each other
- Status updates sent at regular intervals, or triggered by events
- Report generation from live data rather than manual exports
- Routing and approval workflows where rules are clear
The best automation is unremarkable. Nobody notices it, and the process simply stops being a thing anyone has to think about.
Keep a Human in the Loop
Automations fail, and they fail quietly. Design exception paths deliberately so that unusual input surfaces to a person rather than disappearing into a log nobody reads.
Frequently asked questions
Which business processes should be automated first?
Start with high-volume, repetitive and rule-based tasks: data transfer between systems that do not integrate, regular status notifications, report generation from live data, and clear routing or approval workflows. Low-frequency tasks rarely justify the build cost, however tedious they feel.
Should you automate a process before fixing it?
No. If a process depends on undocumented knowledge — someone knowing which customer to escalate to, or which spreadsheet is authoritative — automating it makes the ambiguity faster and harder to remove. Fix the process first, then automate the corrected version.
Is business process automation worth the cost?
For the right processes it clearly is, mainly because it removes variance rather than merely saving time. Automation pays off most where volume is high, the rules are stable and mistakes are expensive. It tends not to pay where the process changes constantly or exceptions dominate.
How do you keep a human in the loop with automation?
Design exception paths deliberately. Most automations fail quietly, so unusual input should surface to a named person with enough context to act, rather than disappearing into a log nobody reads. Alerting without a defined response is the most common reason automations quietly stop being trusted.
Written by Immortify Digital, solutions team at Immortify Digital.
The solutions engineers who map business workflows and connect the systems that already run your operation.
