Create Custom Agents AI-102 Practice Questions (Page 1)

Questions 1–5

1. 1application · medium

   A retail company is building a customer-support agent using the Microsoft Foundry Agent Service. The agent must answer product questions from a private knowledge base and escalate to a human agent when it cannot find a confident answer. The solution must minimize custom code and rely on built-in capabilities. What should you configure?

   • ABuild a custom agent with the Microsoft Agent Framework and implement a fallback intent that triggers an email notification.
   • BProvision an Azure Bot Service, register it with the knowledge base, and use Adaptive Cards to display an escalation form.
   • CDeploy a QnA Maker knowledge base and use the REST API to return a default answer when confidence is low.
   • DCreate an agent in Foundry, attach the knowledge base as a data source, and enable the escalation action to hand off to a human.
   Reveal Solution Discussion Select an answer first
2. 2foundation · easy

   When building an agent that needs to retrieve up-to-date information from a company's internal documents, which additional Azure resource is most important?

   • AAzure DevOps
   • BAzure Monitor
   • CAzure Virtual Network
   • DAzure AI Search
   Reveal Solution Discussion Select an answer first
4. 3foundation · easy

   In a multi-agent solution, what is the role of an orchestrator?

   • ATo train the agents on new data
   • BTo provide a user interface for each agent
   • CTo store all conversation logs for auditing
   • DTo manage the interactions and task assignments among multiple agents
   Reveal Solution Discussion Select an answer first
5. 4expert · hard

   A company has deployed a custom agent using the Microsoft Foundry Agent Service. During testing, they find that the agent's responses are inconsistent: sometimes it provides detailed answers, sometimes very brief ones. They want to optimize the agent for consistent quality. What should they do?

   • ARandomly adjust the temperature parameter for each request to see which works best.
   • BSwitch to a different model that is known to be more verbose.
   • CUse a prompt template with fixed instructions and evaluate the agent against a set of benchmark questions with expected answer lengths.
   • DRemove the system prompt to let the model respond naturally.
   Reveal Solution Discussion Select an answer first
6. 5expert · hard

   A government agency wants to create an agent using the Microsoft Foundry Agent Service to answer citizen questions about public services. The agent must provide accurate information and must not expose sensitive internal data. The agency has a strict budget and wants to minimize ongoing costs. What is the best approach?

   • AConnect the agent to all internal databases and rely on the model's safety filters to prevent leakage.
   • BDeploy the agent on a dedicated GPU cluster to ensure fast response times.
   • CAttach only the approved public documents as a data source and configure the agent to use retrieval augmented generation (RAG) with strict system instructions to avoid internal data.
   • DUse a large language model with a very high temperature to generate diverse responses.
   Reveal Solution Discussion Select an answer first