AI Implementation, Maintenance, and Decommissioning AAIR Practice Questions (Page 4)

Questions 16–20

1. 16application · medium

   After decommissioning an AI system, an organization conducts a post-decommissioning review. The review team is tasked with identifying lessons learned and ensuring compliance with regulatory requirements. Which activity should be part of this review?

   • AEnsure that all stakeholders are satisfied with the decommissioning outcome.
   • BConfirm that the AI system's vendor contract has been terminated.
   • CDocument the decommissioning process, including any deviations from the plan and their reasons.
   • DVerify that all data has been deleted, regardless of retention requirements.
   Reveal Solution Discussion Select an answer first
2. 17foundation · easy

   What is a critical step when executing the decommissioning of an AI system?

   • ADeleting all backups without verification
   • BSecurely deleting or migrating data according to the plan
   • CLeaving the system running to avoid data loss
   • DShutting down the system without notifying users
   Reveal Solution Discussion Select an answer first
4. 18application · medium

   A financial institution is deploying an AI model that approves small business loans. The model must integrate with the core banking system and a customer relationship management (CRM) platform. The deployment team is concerned about data latency and the impact on existing transaction processing. Which deployment strategy should they use to minimize risk to the core banking system?

   • AEmbed the model's scoring logic directly into the core banking system's codebase to ensure tight integration.
   • BDeploy the model as a microservice that is called asynchronously, with results stored in a queue for the CRM to consume.
   • CDeploy the model on a separate server and have the core banking system call it synchronously for each loan application.
   • DUse a nightly batch process to run the model on all loan applications and update the CRM with the results.
   Reveal Solution Discussion Select an answer first
5. 19application · medium

   An e-commerce company is implementing an AI-based recommendation system. Before going live, the data science team must validate that the model performs well across different customer segments and does not introduce bias. The company has a diverse customer base and wants to ensure fair treatment. Which validation approach should they use?

   • AConduct a user acceptance test (UAT) with a small group of customers to gather qualitative feedback on the recommendations.
   • BCompare the model's recommendations against a random baseline to ensure it is better than chance.
   • CEvaluate the model's performance metrics (e.g., precision, recall) on a holdout test set that is stratified by customer segment.
   • DReview the model's feature importance scores to ensure that no protected attributes are used as features.
   Reveal Solution Discussion Select an answer first
6. 20application · medium

   A logistics company is deploying an AI system that optimizes delivery routes. The system will be integrated with the company's fleet management software and will receive real-time traffic data. The company wants to ensure that the system can be updated without interrupting the fleet operations. Which deployment consideration is most important?

   • ADesign the integration to support versioned APIs so that the AI model can be updated without changing the fleet management software.
   • BRequire the fleet management software to be updated whenever the AI model is updated.
   • CUse a monolithic architecture to simplify the deployment process.
   • DDeploy the AI system on the same server as the fleet management software to reduce network latency.
   Reveal Solution Discussion Select an answer first