Imagine someone working in support. When an activity gets flagged it gets assigned to someone. An automated system might take action (eg shadow banking).

Given the automation that worked might process 5000 cases a day.

If the automation can resolve more cases that figure might go up to 8000.

But if there’s an appeal that takes manual review and resolution that might wear up a lot of time stick that the rate drops to 3000.

This means two things:

1. There is an impact on the metric from false positives; and

2. Having human review is still part of the system ultimately. It rather it can always be escalated to such.

Now you might say the worker might be motivated to take the least time consuming action possible even if wrong but an appeal might be escalated above then and further time spent rectifying their mistake still counts against them.