When is a workflow worth automating?
Start with a recurring task and a clear outcome. Here is how to decide where automation can help.
AT A GLANCE
Automation is worth exploring when work repeats, its rules are reasonably clear, and it consumes measurable time or creates errors. Include exceptions, review, and ongoing operation in the decision. A small, well-defined pilot will tell you more than a broad promise.
- Look at exceptions as well as frequency.
- Measure the starting point.
- Start with one complete workflow.
Describe the work first
Choose a specific task, such as turning a website enquiry into a customer record. Describe its trigger, where the data comes from, who checks it, and where the next action appears. “Manage leads” is too broad. “Create a case from a validated form, then notify the owner” describes a workflow you can design.
Look at everyday examples and difficult ones. When colleagues interpret a term differently, clarify the rule first. An unclear workflow remains unclear after automation; it simply produces confusing results more quickly.
Frequency is only part of the picture
A brief daily task and a monthly reconciliation taking several hours can both be reasonable candidates. Record hands-on effort, waiting time, and later corrections separately. This shows which work a system can actually take over.
Frequent exceptions increase development and supervision needs. You do not have to automate all of them. Clear cases can proceed while incomplete ones enter a visible review queue. Someone should own those exceptions, and the right person should hear when work stops. An unattended queue is still a bottleneck, even if the steps before it are automated.
Include the full operating cost
Development is only one part of the investment. Consider provider fees, failure handling, access management, changes to external tools, and training. Deduct the review time that remains in the new process from any estimate of time released.
Avoid treating available working time as an automatic cash saving. If colleagues spend the released time helping customers, that is a capacity improvement whose value needs observing separately. A useful decision model makes uncertainty visible rather than hiding it behind a single confident number. Include the person who will operate the workflow when discussing these assumptions.
Run a small pilot with acceptance criteria
The pilot should handle one complete case from beginning to end. Decide what success means: correct destination, no duplicates, a visible task for its owner, and recoverable state when something fails. Include missing data, repeated submissions, and a temporarily unavailable service in your tests.
Keep human review at first. Compare effort, stalled work, and correction time across a similar volume and mix of cases. If the pilot does not create enough value, that is useful evidence too. You learned at a smaller cost than building the entire system at once. Expand only after the normal path and the exceptions are understood.
A practical first step
Choose one recurring workflow next week and keep a short observation log. Record the case type, hands-on effort, reasons for waiting, and any exception. Use non-identifying reference numbers instead of personal customer data. Make the log simple enough that colleagues will actually maintain it.
At the end of the week, compare common cases. If a stable sequence emerges, you have a useful basis for a conversation. If it does not, focus on clarifying rules and responsibilities. Either outcome moves you towards a better system. You are not looking for the most impressive automation; you are looking for a dependable improvement in the work people do.
A PRACTICAL TEMPLATE
Free workflow observation log
Record one recurring task for a week. The first row is a synthetic worked example, followed by blank rows you can use. Do not include personal customer data.
- Case type
- Hands-on effort
- Reason for waiting
- Exception and correction
Sources and further reading
- Digital and Data Benefits FrameworkUK Cabinet Office · Checked: 2026-10-06
- Measuring the benefits of your serviceGOV.UK Service Manual · Checked: 2026-10-06
Borbély Kristóf
The developer behind Kriside, writing practical guides about web systems, automation, and AI solutions.
