Sustaining AI Transformation and Continuous Improvement CAIPM Practice Questions (Page 7)

Questions 31–35

1. 31foundation · easy

   What is a key consideration when scaling a successful AI solution across an organization?

   • AEnsuring adequate infrastructure and resource allocation to support broader deployment
   • BKeeping the solution isolated to one team
   • CReducing the number of users to maintain performance
   • DAvoiding documentation to speed up replication
   Reveal Solution Discussion Select an answer first
2. 32expert · hard

   A logistics company's AI route-optimization system has been in production for a year. The system has reduced fuel costs by 10%, but the company wants to achieve a 20% reduction. The data science team has proposed several improvements, but each improvement requires a different amount of investment and has uncertain returns. The company has a limited budget and wants to maximize ROI. What is the best approach to decide which improvements to implement?

   • AConduct a portfolio analysis that ranks proposed improvements by expected ROI and risk, then implement the highest-ranked ones within budget.
   • BWait for the next fiscal year to allocate a larger budget for improvements.
   • CChoose the improvement with the lowest upfront cost, regardless of expected return.
   • DImplement all proposed improvements simultaneously to maximize the chance of achieving the 20% target.
   Reveal Solution Discussion Select an answer first
4. 33expert · hard

   A credit union's loan-approval model is subject to fair lending regulations. The model has been in production for two years and is retrained annually. The compliance officer wants to prove that the model has not become biased over time. The data science team has limited storage and cannot keep all historical training data. What should the team do to meet the compliance requirement?

   • ADelete all training data after each retraining to reduce storage costs.
   • BRely on the model's performance metrics from the latest evaluation to prove compliance.
   • CStore model versions, training data summaries, and bias evaluation results for each annual retraining.
   • DKeep only the latest model and its training data to save storage.
   Reveal Solution Discussion Select an answer first
5. 34application · medium

   A retail company's demand-forecasting model has been in production for 18 months. The data science team notices that the model's weekly accuracy has declined from 92% to 84% over the last two months, and the decline correlates with a change in customer purchasing patterns. The team wants to address this without disrupting the current production service. What should the team do first?

   • AInvestigate the root cause by comparing recent data distributions and model performance metrics before deciding on an action.
   • BRetrain the model using the most recent 12 months of data and deploy it immediately to production.
   • CRoll back to the original model version that had 92% accuracy.
   • DSet up automated alerts for data drift and retrain the model monthly on a fixed schedule.
   Reveal Solution Discussion Select an answer first
6. 35application · medium

   A manufacturing company successfully piloted an AI-based predictive maintenance system in one plant. The pilot reduced downtime by 20%. The company wants to expand this success to five other plants, but plant managers are skeptical because they were not involved in the pilot. What should the program manager do first to foster adoption across the other plants?

   • AShare the pilot results and invite plant managers to co-design the rollout plan
   • BImmediately replicate the pilot infrastructure in all plants
   • COffer a financial bonus to plant managers who adopt the system quickly
   • DMandate the use of the system in all plants with a strict rollout deadline
   Reveal Solution Discussion Select an answer first