AI Pilot Execution and Scaled Deployment CAIPM Practice Questions (Page 4)

Questions 16–20

1. 16expert · hard

   A global bank is scaling an AI anti-money-laundering (AML) system. The pilot in one country was successful. The bank must now deploy in multiple countries with different regulatory requirements. The bank also wants to maintain a consistent model across all countries. What is the best scaled deployment strategy?

   • ADevelop a completely separate model for each country
   • BUse a core model with country-specific configuration layers that adjust thresholds and features to meet local regulations
   • CDeploy the exact same model in every country to ensure consistency
   • DDeploy the model only in countries with similar regulations
   Reveal Solution Discussion Select an answer first
2. 17expert · hard

   A credit union is scaling an AI loan-approval model from pilot to production. The model was trained on historical data that underrepresents certain demographic groups. The credit union must comply with fair-lending regulations and wants to avoid disparate impact. What is the most appropriate governance action before full deployment?

   • ADeploy the model as-is because it performed well in the pilot
   • BConduct a bias audit, document the model's limitations, and implement ongoing fairness monitoring in production
   • CRemove demographic data from the model to ensure fairness
   • DRely on the model's explainability report to satisfy regulators
   Reveal Solution Discussion Select an answer first
4. 18application · medium

   A manufacturing firm is scaling an AI predictive-maintenance system from one plant to all plants. Plant managers are skeptical because the pilot was run by a different team. The deployment lead must build trust and adoption. Which action is most effective?

   • ADelay the rollout until all plant managers fully agree.
   • BAssign a central team to manage the system without plant-level involvement.
   • CPresent the pilot results to all plant managers and let them ask questions.
   • DMandate the use of the system in all plants without discussion.
   Reveal Solution Discussion Select an answer first
5. 19application · medium

   An e-commerce company's AI recommendation pilot ran on a single virtual machine with a small database. The scaled deployment will serve 50 times more users and require real-time updates. The infrastructure team must plan for scaling. Which approach should they prioritize?

   • ADeploy the model to multiple on-premises servers without load balancing.
   • BKeep the same virtual machine but increase its size to handle the higher load.
   • CUse a single larger database instance to simplify data management.
   • DMove the model to a serverless inference service and scale the database with read replicas.
   Reveal Solution Discussion Select an answer first
6. 20application · medium

   A manufacturing company successfully piloted an AI quality-inspection system on one production line. The pilot showed a 30% reduction in defect escape. The company now wants to scale to all 20 production lines. What is the most important consideration for the scaled deployment strategy?

   • ADeploy the same model to all lines without additional validation to save time
   • BEnsure each production line has similar data availability and environmental conditions, or adapt the model per line
   • CAsk each line manager to manually adjust the model without oversight
   • DFocus only on scaling infrastructure and ignore model performance
   Reveal Solution Discussion Select an answer first