Measure a representative baseline
Choose one recurring task and record its volume, completion time, error rate and review requirements over a representative period. Include difficult cases as well as easy ones. Agree what counts as a finished item so the before-and-after comparison is fair. Name the person responsible for measuring and the person who will decide whether to proceed.
Count the work that remains
Measure time spent preparing inputs, using the tool, reviewing results, correcting errors and handling exceptions. Include delays caused by access issues or unreliable output. Separate tasks that cannot use the tool from those that can. An attractive demonstration on a handful of easy examples is not a reliable operating baseline.
Work through a conservative example
Suppose a fictional team processes 200 items a week. The current workflow takes 12 minutes per item; the assisted workflow takes eight minutes including review. If 75% of items can use the workflow, the potential capacity released is 200 × 4 ÷ 60 × 0.75 = 10 hours a week. Over 46 working weeks at an illustrative £30 loaded hourly cost, that is £13,800 of annual capacity value before technology, implementation and support costs.
Weekly capacity released = items per week × minutes saved per eligible item ÷ 60 × eligible/adopted share
Annual capacity value = weekly hours × working weeks × loaded hourly cost
Net indicative value = annual capacity value − annual operating costs − relevant implementation costsDistinguish capacity from cash savings
Ten hours released does not automatically reduce payroll by ten hours. The business may use the time for faster responses, more output, better service or a smaller backlog. Record how that capacity will be used and who owns the change. Keep quality and risk measures separate from a financial estimate unless there is a defensible method for assigning them a monetary value.
Set a decision threshold before the pilot
Agree the minimum acceptable quality, reviewer effort, reliability and operating cost. Include licence fees, hosting, maintenance, training and ongoing oversight. Review results with the workflow owner and choose to scale, improve, pause or stop. Repeat the measurement after adoption so an initial improvement is not mistaken for a permanent result.
Practical takeaways
- Compare like-for-like tasks, including exceptions.
- Include review and correction in the assisted time.
- Treat illustrative figures as assumptions, not promises.
- Make a plan for the capacity released.