Automation ROI: how to build a business case without imaginary precision

Establish the workload baseline

Start with annual transaction volume, average active handling time, average waiting time, role mix, rework, and error frequency. Active time and elapsed time are different: automation may improve customer response without releasing an equivalent number of paid hours.

Use a representative sample where direct system data is unavailable. Record the sample period and variability instead of turning a rough recollection into a precise-looking figure.

Use loaded cost carefully

Loaded labour cost can include salary, employer costs, and relevant overhead. Agree which components belong in the decision. Multiplying minutes by an hourly rate estimates capacity value; it does not prove that cash expenditure will fall.

Model more than time saved

Potential benefit may include released capacity, fewer corrections, shorter lead time, reduced leakage, better consistency, or improved auditability. Do not add every benefit to one total unless the relationships are understood; faster handling and lower labour time may describe the same underlying change.

Use conservative, expected, and upper scenarios. A range reveals which assumptions drive the decision and prevents one uncertain number from dominating the case.

Include the whole operating cost

Implementation cost is only one part. Include software subscriptions, usage charges, integration work, testing, security review, training, monitoring, support, maintenance, and the process owner’s time. Allow for the human review that remains by design.

Apply confidence and risk

Reduce expected benefit where data quality, adoption, model accuracy, or integration reliability is uncertain. Identify disqualifying risks separately: a workflow can appear financially attractive and still be inappropriate to automate.

A good business case states what would have to be true for the project to work and what evidence the first release will collect.

Choose the next decision—not the final story

Useful outputs include an acceptable payback range, a maximum responsible first-build budget, a measurable pilot scope, and stop conditions. The calculation should make it easier to say no as well as yes.