Analysts can review every consequential recommendation: Which prior assumption no longer holds?
Stable accuracy does not preserve oversight when exception volume prevents meaningful review.
The question
A retailer uses an AI demand planner, with one analyst reviewing exceptions each morning. Holiday expansion doubles the number of stores and increases exception volume fourfold. Forecast accuracy remains unchanged, but analysts can no longer review every flagged recommendation. This is an EU deployment; applicable EU AI Act obligations apply, while NIST guidance informs internal controls. Which prior assumption no longer holds?
Preparing for AIGP? Take the free 5-min readiness quiz →
- Historical demand data remains relevant to the planner’s forecasts.Historical relevance may persist, but it does not establish that reviewers can examine consequential exceptions.
- The exception threshold still identifies unusual forecasts, so oversight remains operationally adequate.A statistically useful threshold does not make review feasible when its output exceeds human capacity.
- Analysts can review every consequential recommendation before release. ✓Fourfold exception growth invalidates the assumption that human reviewers have sufficient time to provide meaningful oversight.
- The store hierarchy still supports regional reporting and accountability.Reporting structure may remain usable while review workload exceeds available human attention and authority.
The trap
Check actual workload and intervention ability; naming a reviewer is insufficient when volume prevents meaningful examination. How to remember it
Stable accuracy does not preserve oversight when exception volume prevents meaningful review.
How many of these would you get right?
One of 1581 AIGP questions on Certsqill. Take a free five-minute check and see your score per domain — not one number, but which section to open tonight.
Test your AIGP readiness — freeMore Understanding How to Govern AI Deployment and Use questions
- Snapshot incident artifacts and link each record: Which approach best satisfies that constraint? →
- Measure performance for the affected local population: What should the monitoring team do first? →
- Require scoped evaluation and approval before repurposing: What is the defensible next step? →
- All 424 Understanding How to Govern AI Deployment and Use questions →
Part of the Certsqill AIGP question bank · Understanding How to Govern AI Deployment and Use ·
Every answer, right and wrong, comes with its own explanation.
