Design the ALM Process for AI-powered Business Solutions AB-100 Practice Questions (Page 2)

Questions 6–10

1. 6expert · hard

   A company uses Dynamics 365 Customer Service with an AI model that predicts customer sentiment. The model is updated monthly, but the company also uses a custom dataset that is updated in real time. The company wants to ensure that the model is tested with the latest data before deployment, but also wants to minimize the time between data updates and model deployment. What should you do?

   • AUse a fixed dataset for training and ignore real-time updates
   • BDeploy the model without retraining and rely on real-time data at inference
   • CUse a continuous integration pipeline that retrains and evaluates the model whenever the dataset changes, and deploys if performance is acceptable
   • DManually retrain and deploy the model each month
   Reveal Solution Discussion Select an answer first
2. 7expert · hard

   A company uses Dynamics 365 Finance with an AI feature that automates invoice processing. The AI model is updated quarterly, but the company also uses Copilot Studio agents that interact with the same data. The company wants to ensure that updates to the model and agents are coordinated and do not disrupt operations. What should you do?

   • AUse a single environment for all updates to simplify management
   • BDeploy updates only during weekends to minimize impact
   • CUse a shared release calendar and a common staging environment to test model and agent updates together
   • DUpdate the model and agents independently to speed up deployment
   Reveal Solution Discussion Select an answer first
4. 8expert · hard

   A multinational company uses Dynamics 365 Finance with an AI model for credit risk assessment. The model is trained on historical transaction data. They need to deploy a new model version, but the finance team requires a full audit trail of model changes and the data used for training. Additionally, the model must be rolled back within 30 minutes if it produces incorrect results. What should you implement?

   • AA CI/CD pipeline that deploys the model but does not track the training data version.
   • BA model registry that logs every model version, training dataset, and deployment timestamp, with an automated rollback script that can redeploy the previous version within 30 minutes.
   • CA policy to keep the current model and never update it to avoid rollback needs.
   • DA manual process where the data science team documents model changes in a spreadsheet and keeps a backup of the model file.
   Reveal Solution Discussion Select an answer first
5. 9expert · hard

   A company develops a custom AI model for financial forecasting. The model is retrained monthly, but the training data is updated daily. The company wants to ensure that the model is trained on the latest data, but also wants to avoid frequent retraining that could introduce instability. What should you do?

   • ARetrain the model daily to always use the latest data
   • BRetrain the model only when a data scientist manually triggers it
   • CUse a static dataset for training and never update it
   • DUse a scheduled retraining pipeline that runs monthly, but with a data versioning system that allows the model to reference the latest data at inference time
   Reveal Solution Discussion Select an answer first
6. 10application · medium

   A company uses Copilot Studio agents that rely on a shared Azure SQL database for grounding data. The data team updates the database schema quarterly, and the agent team releases new agent versions monthly. They need to ensure that agent updates do not break when the schema changes. What should you recommend?

   • AVersion the database schema and store the schema version in the agent's configuration; release agent updates that are compatible with the current schema version.
   • BDeploy agent updates and schema changes simultaneously without versioning to keep them in sync.
   • CFreeze the database schema and require all data changes to go through the agent release pipeline.
   • DHave the agent team test only against production data to catch schema issues early.
   Reveal Solution Discussion Select an answer first