Create Custom Agents AI-102 Practice Questions (Page 8)

Questions 36–40

1. 36application · medium

   A team has built a customer-support agent using the Microsoft Foundry Agent Service. Before deployment, they want to measure how well the agent handles common customer queries and identify areas for improvement. Which approach should they use?

   • ARun the agent against a test set of real customer queries and compare its responses to expected answers using an evaluation harness.
   • BAsk the development team to manually review the agent's responses to a few sample questions.
   • CDeploy the agent to production and monitor user feedback for a month.
   • DCheck the agent's response time and token usage only.
   Reveal Solution Discussion Select an answer first
2. 37foundation · easy

   Which component is typically defined when creating an agent in the Microsoft Foundry Agent Service?

   • AThe agent's system prompt or instructions
   • BThe programming language used to write the agent
   • CThe physical server location for the agent
   • DThe billing plan for the underlying virtual machines
   Reveal Solution Discussion Select an answer first
4. 38expert · hard

   A media company is building a multi-agent system to generate news articles. A research agent gathers facts, a writer agent drafts the article, and an editor agent reviews for quality. The system must allow human editors to approve articles before publication. What is the best orchestration pattern?

   • AUse a pipeline where the research agent passes facts to the writer, the writer passes the draft to the editor, and the editor's output is sent to a human approval step before publication.
   • BUse a single agent that performs research, writing, and editing in one step.
   • CHave the writer agent directly publish the article after drafting, with the editor agent running asynchronously.
   • DRun all three agents in parallel and then automatically publish the best result.
   Reveal Solution Discussion Select an answer first
5. 39expert · hard

   A global bank is implementing a multi-agent system for loan processing. One agent verifies identity, another checks credit history, and a third approves or rejects the loan. The system must comply with regional data residency regulations, meaning customer data must stay within specific geographic regions. The agents are built with the Microsoft Agent Framework. What is the best way to design the orchestration?

   • AUse a single agent that performs all three steps to avoid cross-region data movement.
   • BUse a single global orchestrator that sends all data to a central region for processing.
   • CRun all agents in one region and use data masking to anonymize customer data before processing.
   • DDeploy separate agent instances in each region and configure the orchestrator to route requests to the appropriate regional instance based on the customer's location.
   Reveal Solution Discussion Select an answer first
6. 40application · medium

   A real estate company wants to build an agent that helps clients find properties based on their preferences. The agent should ask clarifying questions and provide personalized recommendations. Which use case and resource configuration best fits this requirement?

   • AA search engine that returns all properties matching a keyword query.
   • BA conversational agent that uses a property database as a data source and is built with the Microsoft Foundry Agent Service.
   • CA recommendation system that uses collaborative filtering on past user behavior.
   • DA form-based tool that collects preferences and runs a SQL query.
   Reveal Solution Discussion Select an answer first