Calculator Library / AI Operations
Enter a measured baseline and your assumptions to see the monthly benefit, payback month, and 36-month cash for one AI use case, with a sensitivity table that shows which assumptions matter most.
Tool
Net = labor saving + error saving − running cost
Enter the baseline and your assumptions for one AI use case. The calculator gives the monthly benefit, the payback month, the cumulative net cash over 36 months with a ramp-up period, and a sensitivity table that shows which assumptions matter most. Load the example to see a complaint-classification case.
Everything runs in your browser. Nothing is sent to a server, and entries are not saved between visits. The example is made-up data. Use your own measured baseline.
Result
Picture
Enter the case to draw the chart.
Share handled and time saved are capped at 100%. A dash means the case does not pay back in 36 months.
Instructions
The calculator turns a handful of measurable inputs into a business case for one AI use case. It works out how many items the AI handles, the net minutes saved after human review, the monthly labor and error savings, the net benefit after running costs, the payback month with a ramp-up period, and the cumulative net cash over 36 months. A sensitivity table shows which assumptions move the result most.
It matches the worked example in the AI Use-Case Selection and Business Case guide. To choose which use cases to cost, start with the AI Use-Case Prioritization Matrix.
| Line | Formula |
|---|---|
| Items handled | Items per month × share handled |
| Net minutes saved per item | Minutes today × share of time saved − review minutes |
| Labor saving per month | Items handled × net minutes ÷ 60 × hourly cost |
| Error saving per month | Items handled × (error rate today − error rate with AI) × cost of an error |
| Net per month at full benefit | Labor saving + error saving − running cost |
| Cumulative net in month m | − one-time cost + sum over months of [(labor + error saving) × ramp factor − running cost], where the ramp factor is the smaller of 1 and m ÷ months to full benefit |
| Payback month | The first month in which the cumulative net is zero or more |
The pre-loaded example is complaint classification at an illustrative plant: 600 complaints a month, 12 minutes of affected handling time each, 80% handled, 60% time saved, 2 review minutes, $38 an hour, $450 a month to run, $18,000 to set up, and misroutes falling from 4% to 3% at $60 each, with a 3-month ramp. The net benefit is $1,419 a month at full benefit, payback is in month 15, and net cash after 36 months is $31,208.
It is the first month in which cumulative net cash, after the one-time cost, the running costs, and the ramp-up of benefits, reaches zero. Because benefits ramp up gradually, payback is later than the one-time cost divided by the monthly net benefit.
Use pilot data if you have it: the share of items the AI processes to the standard you need without a person redoing them. If you have no pilot yet, use a cautious figure and look at the sensitivity table to see how much the answer depends on it.
Because most AI uses need a person to check the output at first. The net time saved is the time removed from the task minus the time spent checking. A case that ignores review time overstates the saving.
Enter the share of items that have a costly error today and with the AI in place, after review, and the average cost of one error. If the AI makes more errors than people, the error saving will be negative, which reduces the net benefit.
No. The calculator runs entirely in your browser, nothing is sent to a server, and your entries are not saved when you leave the page.