Sustaining AI Transformation and Continuous Improvement CAIPM Practice Questions (Page 6)

Questions 26–30

1. 26foundation · easy

   What is a common approach to assessing the long-term value of an AI initiative?

   • ATracking KPIs over time and adjusting strategy based on observed trends
   • BComparing against competitors without context
   • CEvaluating only the initial implementation cost
   • DRelying solely on anecdotal feedback from users
   Reveal Solution Discussion Select an answer first
2. 27expert · hard

   A bank has an AI credit-scoring model that is retrained every six months. The model's performance has been stable, but the bank's regulator now requires that the model be able to explain its decisions. The model is a complex ensemble that is difficult to interpret. The data science team wants to replace it with a simpler, interpretable model, but the simpler model has slightly lower accuracy. What should the program manager do?

   • AKeep the complex model and use a separate explanation tool to generate explanations
   • BKeep the complex model and argue that its accuracy justifies its use
   • CRetrain the complex model more frequently to improve its explainability
   • DReplace the complex model with the simpler model and document the trade-off in accuracy
   Reveal Solution Discussion Select an answer first
4. 28application · medium

   An insurance company's fraud-detection model is monitored with a dashboard that tracks precision, recall, and the number of flagged cases. Recently, the number of flagged cases has increased sharply, but the fraud rate among flagged cases has dropped. The team suspects the model is now over-flagging. What should the team do to validate this suspicion?

   • AIgnore the change because the model is still detecting more fraud than before.
   • BRetrain the model immediately with the latest data to improve precision.
   • CIncrease the model's decision threshold to reduce the number of flagged cases.
   • DExamine the distribution of input features in recent claims compared to the training data.
   Reveal Solution Discussion Select an answer first
5. 29expert · hard

   An insurance company uses an AI model to set premiums. The model is updated quarterly, and each update must be approved by the compliance team. The data science team wants to move to continuous deployment to improve responsiveness. The compliance team is concerned about the ability to audit changes. What is the best approach to satisfy both teams?

   • AImplement continuous deployment but require automated compliance checks and full audit logging for every model version
   • BAllow the data science team to deploy changes without compliance review if the model's accuracy improves
   • CMaintain quarterly updates to preserve the current audit process
   • DDeploy changes to production only after a full manual compliance review each time
   Reveal Solution Discussion Select an answer first
6. 30foundation · easy

   Which iterative methodology is commonly applied to AI systems to enable continuous improvement?

   • APDCA (Plan-Do-Check-Act)
   • BWaterfall development
   • CBig Bang deployment
   • DFirefighting approach
   Reveal Solution Discussion Select an answer first