AI Pilot Execution and Scaled Deployment CAIPM Practice Questions (Page 8)

Questions 36–40

1. 36application · medium

   After deploying an AI churn-prediction model in production, the data science team notices that the model's accuracy is declining over time because customer behavior is changing. What is the most appropriate post-deployment action?

   • AImplement continuous monitoring and a retraining pipeline that triggers when performance drops below a threshold
   • BIgnore the decline because the model was accurate at deployment
   • CFreeze the model to ensure consistency
   • DRetrain the model manually once a year
   Reveal Solution Discussion Select an answer first
2. 37expert · hard

   A media company's AI content personalization pilot ran on a batch processing pipeline that updated recommendations nightly. The scaled deployment requires near-real-time recommendations for millions of users, but the infrastructure budget is fixed. The team must choose an architecture. Which option best balances performance and cost?

   • AUse a stream-processing platform with auto-scaling to handle variable load and keep costs aligned with usage.
   • BKeep the nightly batch pipeline and increase the batch frequency to every hour.
   • CUse a single powerful server with in-memory caching to serve all users.
   • DProvision a large cluster of always-on servers to handle peak traffic and minimize latency.
   Reveal Solution Discussion Select an answer first
4. 38expert · hard

   A media company is scaling an AI content-personalization service. The pilot used a fixed cluster of 10 servers. The full deployment will have unpredictable traffic spikes during major events. The company has a limited budget and wants to avoid over-provisioning. What is the best infrastructure scaling strategy?

   • AUse a small fixed cluster and accept downtime during spikes
   • BUse auto-scaling with a minimum and maximum cluster size, and configure policies to scale based on traffic metrics
   • CProvision a large fixed cluster to handle peak traffic at all times
   • DUse manual scaling and add servers only when the team remembers
   Reveal Solution Discussion Select an answer first
5. 39foundation · easy

   What is the goal of continuous optimization after AI deployment?

   • ATo reduce the number of users of the AI solution
   • BTo improve the AI solution's performance and business value over time
   • CTo keep the AI solution exactly as it was at launch
   • DTo eliminate the need for monitoring
   Reveal Solution Discussion Select an answer first
6. 40application · medium

   An online retailer deployed an AI pricing model. After three months, the model's revenue impact has plateaued. The data science team suspects the model is not adapting to changing market conditions. What should the team do to sustain value?

   • AIncrease the model's complexity to capture more market nuances.
   • BReplace the model with a simpler rule-based system to reduce costs.
   • CSet up a regular retraining schedule and monitor for concept drift using live data.
   • DStop monitoring the model because it has already proven its value.
   Reveal Solution Discussion Select an answer first