Workflows & automation · JNAGA perspective
Should you fix a process before automating it?
The best automation candidate is not always the task people dislike most. It is work whose purpose and boundaries are clear enough to change safely.
Published 25 September 2026
Follow the work as it happens
Write down a real case from beginning to end. Who starts it, which information is needed, where does it pause and who decides an exception? The formal process diagram may miss the email thread or spreadsheet that makes the service possible in practice.
If two teams use different definitions for the same status, automating the handover can preserve the disagreement at greater speed. Clarify the shared meaning first.
Separate judgement from repetition
Repeated copying, reminders and predictable checks may be suitable for automation. Ambiguous requests, sensitive decisions and unusual exceptions may need a person. Design an explicit route between the two rather than pretending every case follows the happy path.
Ask what evidence an operator needs to trust the automated result and how they will reverse or correct it. A workflow that cannot be explained or recovered is expensive to own.
Improve the process and the system together
A small simplification may remove a step entirely. Where automation still helps, test it against real cases and assign an owner for exceptions, monitoring and change. The aim is less avoidable effort and better service, not a higher count of automated actions.
Imagine an approval request that usually contains a complete form. Sometimes a document is missing and the approver must decide whether work can proceed.
Automate the completeness check and routine routing, but keep the missing-document decision with an accountable person. The exception path is part of the design, not a failure of automation.