Demonstrate a tested manual inspection fallback: Which evidence supports the safest immediate control?
A tested manual fallback allows the affected function to stop while production retains a controlled inspection process.
The question
A factory’s AI flags defective parts, but a software update causes unexplained false alerts. The team must decide whether production can continue while investigating. Which evidence supports the safest immediate control?
Preparing for AIGP? Take the free 5-min readiness quiz →
- Compare the updated model with its prior accuracy score.A score comparison may identify performance change, but it does not prove production can continue safely during unresolved failures.
- Demonstrate a tested manual inspection fallback. ✓A tested fallback preserves quality-control operations while the affected AI function is suspended and the unexplained alerts are investigated.
- Wait for the supplier’s next release before changing production procedures.Waiting leaves the affected function operating without containment and provides no evidence that the next release resolves this specific failure.
- Ask operators to monitor alerts more carefully during each shift.Additional attention may help temporarily, but it leaves the unreliable function active without a validated alternative process.
The trap
The key evidence is a tested alternative process that works while the affected AI function is contained. How to remember it
A tested manual fallback allows the affected function to stop while production retains a controlled inspection process.
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
- Run a tabletop of detection: Which evidence is most useful? →
- Reassess the updated version against the actual workflow: Which evidence should be required before continued →
- Track proxy signals and later verified outcomes: Which approach is most appropriate? →
- 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.