AI Strategy and Adoption Roadmap Design CAIPM Practice Questions (Page 9)

Questions 41–45

1. 41application · medium

   An insurance company has had an AI claims-triage system in production for six months. The system's accuracy has declined from 88% to 81% because customer behavior and claim patterns have changed. The business team wants to keep the system running without interruption. What is the most appropriate mechanism for continuous improvement?

   • AFreeze the model to maintain consistent decisions and avoid further decline
   • BSet up automated performance monitoring with alerts and a monthly retraining pipeline
   • CReplace the model with a rule-based system to eliminate drift
   • DSchedule a quarterly model retraining cycle using the latest claims data
   Reveal Solution Discussion Select an answer first
2. 42expert · hard

   An insurance company is deploying an AI system for claims processing. The system uses a complex model that is not fully interpretable. The regulator requires that the company be able to explain any adverse decision to a customer. The company has a strong data science team but limited experience with regulatory compliance in AI. What is the most appropriate governance approach?

   • ADefer the deployment until a fully interpretable model can be developed.
   • BReplace the complex model with a simpler, interpretable model, even if it reduces accuracy.
   • CImplement a two-tier review process where the AI flags adverse decisions for human review, and the human reviewer documents the rationale for the decision.
   • DUse the complex model but do not document the rationale for adverse decisions, relying on the model's overall accuracy.
   Reveal Solution Discussion Select an answer first
4. 43application · medium

   A media company has deployed an AI content-recommendation system. After three months, user engagement has plateaued. The data team suspects the model is no longer learning from new user behavior. The product manager wants to improve engagement. What is the most appropriate action?

   • AIncrease the frequency of A/B tests without changing the model
   • BReplace the recommendation algorithm with a completely different approach
   • CRetrain the model with recent user interaction data and monitor engagement metrics
   • DKeep the model unchanged to maintain consistent recommendations
   Reveal Solution Discussion Select an answer first
5. 44foundation · easy

   What is the primary purpose of defining KPIs for AI adoption?

   • ATo measure the success and impact of AI initiatives
   • BTo replace the need for project management
   • CTo eliminate the need for stakeholder feedback
   • DTo guarantee model accuracy
   Reveal Solution Discussion Select an answer first
6. 45application · medium

   A healthcare provider is planning a two-year AI roadmap. The first year focuses on a pilot for patient readmission prediction, and the second year aims to scale it across all departments. The finance team has approved a fixed budget. What is the most effective way to allocate resources across the roadmap?

   • AAllocate 50% to the pilot and 50% to scaling, with a contingency reserve for unexpected costs in either phase.
   • BAllocate 70% of the budget to the pilot phase and 30% to scaling, ensuring the pilot is fully funded to prove value.
   • CAllocate 90% to the pilot and 10% to scaling, because the pilot is the most critical phase.
   • DAllocate 40% to the pilot and 60% to scaling, reserving the majority for the larger deployment effort.
   Reveal Solution Discussion Select an answer first