Regulators increasingly ask not only whether a human reviews automated recommendations, but how often reviewers disagree, how long they take and whether review is genuine. Human review measures it: recommendations received, reviewed and overridden, the agreement rate, the median review time, and the share of very fast reviews that may signal rubber-stamping, broken down by reviewer group.
The figures come from recorded events. Where the organisation has not instrumented a decision, the screen says so plainly instead of showing a reassuring blank.
On this screen you can
- Compare reviewer groups on agreement rate and review time.
- Set a threshold for what counts as a fast review and see how many cross it.
- Trace any figure back to the events it was calculated from.
