Create Custom Agents AI-102 Practice Questions (Page 6)

Questions 26–30

1. 26application · medium

   A legal firm wants to create an agent that helps lawyers search case law and legal precedents. The agent must use a private Azure AI Search index that contains licensed legal documents. The firm wants to build the agent using the Microsoft Foundry Agent Service. Which resource configuration is required?

   • AA public website hosting the documents, and the agent uses web search to find them.
   • BAn Azure AI Search resource with the legal documents indexed, and a connection from the Foundry agent to that search index.
   • CAn Azure SQL database with the documents stored as text, and a direct connection from the agent to the database.
   • DA Power Automate flow that retrieves documents from SharePoint and passes them to the agent.
   Reveal Solution Discussion Select an answer first
2. 27foundation · easy

   Which Azure resource is typically required to give an agent access to a large language model?

   • AAzure Blob Storage
   • BAzure AI Search
   • CAzure Functions
   • DAzure OpenAI Service
   Reveal Solution Discussion Select an answer first
4. 28application · medium

   An e-commerce company is building a multi-agent system: one agent handles product recommendations, another processes returns, and a third tracks shipments. The system must allow users to interact with all three agents through a single chat interface. Which architecture should they use?

   • AA single agent that combines all three functions into one model.
   • BThree separate chat interfaces, one for each agent.
   • CA front-end agent that routes user messages to the appropriate specialized agent based on intent.
   • DA rule-based system that maps keywords to agents without using AI.
   Reveal Solution Discussion Select an answer first
5. 29expert · hard

   A financial advisory firm is building an agent with the Microsoft Agent Framework that must analyze market data, generate investment recommendations, and execute trades. The firm requires that all recommendations be logged and that trades are only executed after a human trader approves them. Which implementation should they choose?

   • ALet the agent execute trades automatically to avoid delays.
   • BLog only the final trade, not the recommendations.
   • CUse a workflow that logs every recommendation, pauses for human approval, and then executes trades through a separate tool.
   • DUse a single tool that both recommends and executes trades without separation.
   Reveal Solution Discussion Select an answer first
6. 30expert · hard

   A healthcare startup is building a complex agent using the Microsoft Agent Framework to assist doctors with patient data retrieval and appointment scheduling. The agent must integrate with an EHR system that only supports OAuth 2.0 and has rate limits. The solution must handle multiple concurrent requests without exceeding the rate limit. What is the best approach?

   • ACache all patient data locally to avoid calling the EHR system frequently.
   • BUse a single shared API key for all requests to simplify authentication.
   • CLimit the agent to one concurrent request to avoid hitting the rate limit.
   • DImplement a tool that uses OAuth 2.0 to authenticate and includes a retry policy with exponential backoff to handle rate limits.
   Reveal Solution Discussion Select an answer first