Design the ALM Process for AI-powered Business Solutions AB-100 Practice Questions (Page 5)

Questions 21–25

1. 21application · medium

   A company has a Copilot Studio agent that uses custom connectors to call an internal API. The API team frequently updates the API endpoints. The agent team wants to test new agent versions with the latest API changes without affecting the production agent. What should you do?

   • AUse the same connector for both environments but rely on feature flags to toggle API endpoints.
   • BUpdate the production connector directly and test the agent in production.
   • CAsk the API team to freeze the API until the agent is stable.
   • DCreate a separate development environment with its own connector configuration pointing to a test API endpoint.
   Reveal Solution Discussion Select an answer first
2. 22application · medium

   A team manages Microsoft Foundry Agents that use a shared vector index for retrieval-augmented generation. The index is rebuilt weekly. The team wants to ensure that agent updates are tested with the latest index before production. What should you do?

   • ATest the agent only with a sample of the index
   • BInclude the index version in the agent's deployment pipeline and test with the latest index in a staging environment
   • CDeploy the agent to production and then rebuild the index
   • DUse a static index that is never updated
   Reveal Solution Discussion Select an answer first
4. 23expert · hard

   A company uses Copilot Studio agents and Microsoft Foundry Agents that share a common data source. The data source is updated in real time, and the agents need to reflect changes immediately. However, the data team requires a 24-hour delay before new data is available to agents for compliance reasons. How should you design the ALM process?

   • ACreate a staging data store that receives real-time updates but only publishes to the agents after a 24-hour compliance hold.
   • BUpdate the agents' code to ignore new data for 24 hours.
   • CDelay all data updates by 24 hours at the source.
   • DAllow agents to access the real-time data directly and rely on the data team to audit later.
   Reveal Solution Discussion Select an answer first
5. 24application · medium

   A customer service team uses Dynamics 365 Customer Service with an AI agent that suggests responses. The agent is updated frequently based on feedback. They want to test new agent versions with a subset of real customer interactions before full deployment. What should you configure?

   • AA full deployment of the new agent version to all users immediately.
   • BA manual review of all suggested responses before they are sent.
   • CA test environment that uses only synthetic customer interactions.
   • DA staged rollout that routes a small percentage of live interactions to the new agent version while monitoring performance.
   Reveal Solution Discussion Select an answer first
6. 25application · medium

   A company develops a custom AI model for supply chain demand forecasting. The model is retrained weekly with new sales data. The company wants to automate the retraining, validation, and deployment process. What should you implement?

   • AAn Azure Machine Learning pipeline that retrains, evaluates, and registers the model, with a separate release pipeline for deployment
   • BA single script that retrains and deploys the model in one step
   • CA scheduled job that deploys the model without retraining
   • DA manual process where data scientists retrain and deploy the model
   Reveal Solution Discussion Select an answer first