We can generate faster than we can review.
Checking can't keep up
People and AI produce more than the few people who can judge it have time to check. Reviews get shorter, the queue gets longer, and the load falls on a handful of senior people. The system can look fine until a serious error gets through or one of the overloaded reviewers leaves.
- The queue of work waiting for review grows faster than it clears.
- The hardest cases get the same rushed look as the easy ones.
- Everyone routes around one overloaded expert.
Who: the people who review work in this workflow. Count: items waiting for review each Monday for two weeks. Look at: how long three recent reviews actually took, next to how long a careful check would take. Ask: "How much can you review in a week before it turns into approving without reading?" You're probably looking at this pattern if the queue is growing and reviews are much shorter than a careful check. Probably not if the queue is steady and reviewers have time.
Expertise spent on low-value work. In that pattern, the expert's time goes to preparation. In this pattern, it goes to reviewing, and there isn't enough of it.
Agree that the person checking can pause new output at any time, and that the pause is honored without argument.
Count what is open at the review stage. Set a number. Agree that nothing new starts until something clears.
Start with two triggers. The model flags when it is unsure, and when the decision is hard to reverse.
People miss more errors in automated work when they are under heavy load (Parasuraman and Manzey, 2010). Established.