Design the ALM Process for AI-powered Business Solutions AB-100 Practice Questions (Page 6)

Questions 26–30

1. 26application · medium

   A company is developing a customer-support Copilot agent that uses a custom AI model trained on historical ticket data. The data science team frequently updates the training dataset, while the Copilot team iterates on the agent's prompts and actions. The company needs to ensure that changes to the data, model, and agent are tested together before production deployment. What should you implement?

   • ASeparate pipelines for the model and the agent, each deploying to production independently
   • BA single pipeline that deploys the model to production first, then deploys the agent
   • CA shared Azure DevOps pipeline that builds the model, packages the agent, and runs integration tests on a staging environment
   • DA manual process where the data team updates the dataset and the Copilot team updates the agent without coordination
   Reveal Solution Discussion Select an answer first
2. 27application · medium

   A data science team uses a custom AI model that depends on a dataset that is updated by an external vendor. The vendor provides a new dataset version monthly. The team needs to ensure that the model is retrained with the new data and deployed without manual intervention. What should you implement?

   • AA policy to skip retraining until the model performance degrades.
   • BA scheduled job that retrains the model with the old dataset.
   • CAn automated pipeline that detects the new dataset version, retrains the model, and deploys it after validation.
   • DA manual process where a data scientist downloads the dataset and retrains the model.
   Reveal Solution Discussion Select an answer first
4. 28application · medium

   A data science team trains custom AI models using datasets that are updated weekly. They need to track which model version was trained on which dataset version to ensure reproducibility. What should you implement?

   • AA manual spreadsheet that tracks model and dataset versions.
   • BA model registry that records the dataset version, model version, and training parameters for each run.
   • CA shared folder where all model files and datasets are stored without versioning.
   • DA naming convention that includes the date in the model file name.
   Reveal Solution Discussion Select an answer first
5. 29application · medium

   A company uses Dynamics 365 Customer Service with a Copilot Studio agent that handles refund requests. The agent's behavior is updated based on policy changes. They need to ensure that the updated agent complies with new refund policies before going live. What should you do?

   • AHave a business analyst manually test the agent in production.
   • BInclude policy validation tests in the CI/CD pipeline that verify the agent's responses against the new policy.
   • CSkip testing because the policy change is minor.
   • DDeploy the agent and monitor customer feedback for issues.
   Reveal Solution Discussion Select an answer first
6. 30expert · hard

   A company uses Dynamics 365 Finance and Customer Service. They have an AI model that predicts customer payment behavior, used in both apps. The finance team requires a strict change approval process, while the customer service team needs rapid updates to respond to market changes. They need a single ALM process that satisfies both. What should you do?

   • AImplement a tiered approval process where changes are classified by risk, with high-risk changes requiring finance approval and low-risk changes following a faster path.
   • BUse the finance team's strict approval process for all changes.
   • CUse the customer service team's rapid process for all changes.
   • DCreate two separate ALM processes, one for each app.
   Reveal Solution Discussion Select an answer first