Create Custom Agents AI-102 Practice Questions (Page 3)

Questions 11–15

1. 11application · medium

   A healthcare organization is implementing a multi-agent solution where a triage agent collects patient symptoms, a scheduler agent books appointments, and a follow-up agent sends reminders. The solution must support multiple concurrent users and allow agents to hand off to each other. What should you implement?

   • AUse the Microsoft Agent Framework to define each agent and configure orchestration with handoff actions between them.
   • BDeploy each agent as a separate Azure Function and use HTTP calls to invoke the next agent.
   • CCreate a single agent with all three roles in its system prompt and let it decide which role to play.
   • DUse the Foundry Agent Service to create a single agent that can call all three APIs directly.
   Reveal Solution Discussion Select an answer first
2. 12foundation · easy

   Which technique is commonly used to evaluate the quality of an agent's responses?

   • ADeploying the agent to production immediately
   • BComparing responses against a set of expected answers or ground truth
   • CRunning the agent on a faster virtual machine
   • DIncreasing the token limit of the model
   Reveal Solution Discussion Select an answer first
4. 13application · medium

   A marketing team wants to create an agent that generates social media posts based on the company's brand guidelines. The guidelines are stored in a PDF file. The team wants to use the Microsoft Foundry Agent Service. What should they do to ensure the agent follows the guidelines?

   • AUse the agent's default behavior and hope it follows general marketing best practices.
   • BTrain a custom model on the brand guidelines.
   • CCopy the entire PDF content into the system prompt.
   • DUpload the PDF to a data source and attach it to the agent as grounding data.
   Reveal Solution Discussion Select an answer first
5. 14foundation · easy

   What is the primary purpose of the Microsoft Foundry Agent Service?

   • ATo provide a managed environment for creating, configuring, and deploying AI agents
   • BTo replace all other Azure AI services with a single unified API
   • CTo train custom large language models from scratch
   • DTo provide a data storage solution for agent conversation logs
   Reveal Solution Discussion Select an answer first
6. 15expert · hard

   A manufacturing company is building an agent with the Microsoft Agent Framework that must handle a complex workflow: when a sensor reading exceeds a threshold, the agent must automatically create a maintenance ticket, notify the shift supervisor, and order a replacement part if the part is in stock. The workflow must be resilient to API failures and allow for human approval before ordering parts. Which approach should the team take?

   • AUse a simple prompt that instructs the agent to perform all steps without any error handling.
   • BLet the agent call all APIs in parallel without retries and order parts automatically to minimize delay.
   • CImplement the agent with a stateful workflow that includes retry logic for API calls and a human-in-the-loop approval step before ordering parts.
   • DHardcode the entire workflow in a single function that the agent calls once.
   Reveal Solution Discussion Select an answer first