Microsoft AB-100 Online Practice Questions and Exam Preparation

Microsoft AB-100 Online Questions & Answers

• Question 31:

  HOTSPOT

  A company uses Azure OpenAI models that use grounding data from Microsoft Fabric for agents. The models are fine-tuned by using proprietary datasets.

  You need to design a governance solution that meets the following requirements:

  1. Restricts access to the grounding data to only assigned roles

  2. Restricts model fine-tuning to only the AI engineering team.

  What should you include in the design? To answer, select the appropriate options in the answer area.

  NOTE: Each correct selection is worth one point.

  

  

  ---

  The completed hotspot maps grounding-data access restriction to Microsoft Purview access policies and fine-tuning restriction to role-based access control in Microsoft Foundry. The answer image separates data-governance control from model-operation authorization. Purview helps govern access to protected enterprise data, while Foundry RBAC controls who can perform model and project operations such as fine-tuning. Content Safety and Monitor do not restrict access to data or fine-tuning permissions.

• Question 32:

  You need to design a Microsoft 365 Copilot solution to optimize employee productivity. The solution must meet the following requirements:

  1. Ensure that the employees can query content stored in a subset of Microsoft SharePoint Online sites and in Teams by using natural language-based prompt actions.

  2. Ensure that employees receive contextually relevant responses in Microsoft 365 Copilot.

  What should you include in the design?

  A. Build a Microsoft Power Automate desktop flow to read the SharePoint content and post the responses to Teams.
  B. Modify SharePoint settings.
  C. Create a custom REST API that crawls the SharePoint content.
  D. Configure Microsoft Graph access.
  D. Configure Microsoft Graph access.

  ---

  Explanation

  Configuring Microsoft Graph access is required because Microsoft 365 Copilot uses Graph-connected Microsoft 365 content and permissions to answer user queries. The requirement is to let employees query a subset of organizational content while respecting access. A standalone search or unmanaged data copy would bypass the permission model. For Microsoft 365 Copilot grounding, make the content available through Microsoft Graph and preserve the user's existing permissions.

• Question 33:

  A company has a Microsoft Foundry agent that summarizes customer feedback and recommends products to customers. The agent references data from multiple knowledge sources.

  Users report that the agent response time is slow.

  Telemetry data shows that the agent frequently reaches its token usage limit.

  You need to recommend a solution to reduce token usage without degrading the quality of the generated responses.

  What should you recommend?

  A. Chunk documents during indexing.
  B. Reduce the number of knowledge sources used by the agent.
  C. Reconfigure the prompts to limit the amount of retrieved content from the knowledge sources.
  D. Lower the maximum token usage limit for the responses.
  C. Reconfigure the prompts to limit the amount of retrieved content from the knowledge sources.

  ---

  Explanation

  Limiting the amount of retrieved content in prompts reduces token usage while preserving the quality of the most relevant context. The telemetry shows that the agent reaches token limits, so the issue is too much context being sent to the model. Removing knowledge sources could reduce quality, and lowering the token limit may truncate useful answers. Tune retrieval scope before removing sources or reducing output capacity.

• Question 34:

  A company deployed several Copilot Studio agents. You need to identify which agents have low engagement, high escalation, and poor answer quality so that the team can tune topics and knowledge sources.

  Which two activities should you include in the monitoring process? Each correct answer presents part of the solution.

  A. Review agent analytics and performance metrics.
  B. Delete low-performing topics before reviewing transcripts.
  C. Interpret telemetry and user feedback to identify tuning opportunities.
  D. Turn off all knowledge sources and rely on the model's general knowledge.
  E. Replace every agent with a custom model before collecting metrics.
  A. Review agent analytics and performance metrics.
  C. Interpret telemetry and user feedback to identify tuning opportunities.

  ---

  Explanation

  Agent analytics and performance metrics show which agents have low engagement, high escalation, or poor answer quality, and telemetry plus user feedback helps interpret why those patterns occur. The requirement is to identify tuning opportunities across several deployed agents. Raw transcripts alone can help investigate individual sessions, but aggregate metrics and feedback make the weak agents visible first. Start with analytics, then drill into the causes.

• Question 35:

  HOTSPOT

  A company plans to implement an AI solution that will contain a Microsoft Copilot Studio agent and a Microsoft Foundry agent. The solution will be stored in a source code repository.

  You need to recommend a deployment method for each agent. The solution must meet the following requirements:

  1. A test environment must be used before a deployment to production.

  2. Production must be isolated from development and testing.

  3. The deployment must be repeatable and fully automated.

  4. The solution must NOT require manual intervention.

  Which deployment method should you recommend for each agent? To answer, select the appropriate options in the answer area.

  NOTE: Each correct selection is worth one point.

  

  

  ---

  The completed deployment design should use a Microsoft Power Platform deployment pipeline for Copilot Studio and an automated source-controlled deployment such as an Azure DevOps pipeline with infrastructure as code for Microsoft Foundry. The answer area shows the need to separate low-code Power Platform promotion from code-first Azure deployment. Copilot Studio agents move through environments as solution components, while Foundry resources are usually deployed from a repository through CI/CD and declarative templates. The requirement for no manual intervention is the key deployment clue.

• Question 36:

  A Copilot Studio agent must collect missing purchase-order values from a user, validate the values, and then run a backend approval process that moves data between services. The backend process must be reusable and trackable outside the conversation.

  What should you include?

  A. An agent flow or Power Automate flow invoked as a tool
  B. A public website knowledge source
  C. A model router only
  D. A topic name that includes every purchase-order field
  A. An agent flow or Power Automate flow invoked as a tool

  ---

  Explanation

  An agent flow or Power Automate flow invoked as a tool fits because the agent must collect values, validate them, and then run a reusable backend approval process. The flow provides deterministic automation while the agent handles conversation and data collection. A topic alone would not be the best reusable backend process boundary. Use tools or flows for business process execution behind an agent.

• Question 37:

  A company wants to create a shared prompt library for AI-powered business solutions. The library must promote reusable formats, consistent responses, governance, and version control.

  Which approach should you recommend?

  A. Allow each department to write unrelated prompts with no review process.
  B. Store prompt text only in private chat history.
  C. Define standardized prompt templates with owners, test criteria, and version control.
  D. Ask makers to keep prompts vague so each model can decide the format.
  C. Define standardized prompt templates with owners, test criteria, and version control.

  ---

  Explanation

  Standardized prompt templates with owners, test criteria, and version control create a governed prompt library. Templates promote consistent response formats, while ownership and testing keep prompts maintainable and reliable. A shared folder of ad hoc prompts would not provide governance or release history. Treat reusable prompts as governed assets with versioning, not informal text snippets.

• Question 38:

  DRAG DROP

  Which Copilot Studio analytics metrics should you recommend to assist the executives with their specific responsibilities? To answer, drag the appropriate metrics to the correct executives. Each metric may be used once, more than once, or not at all. You may need to drag the split bar between panes or scroll to view content.

  NOTE: Each correct selection is worth one point.

  Select and Place:

  

  

  ---

  The answer-area image maps the CFO's concern about Copilot Studio credit usage to Effectiveness and maps the CTO's concern about poor feedback on AI agent responses to Satisfaction. Effectiveness analytics include conversation outcomes such as resolved, escalated, and abandoned sessions, which directly supports the CFO's need to compare abandoned sessions with resolved sessions and identify inefficient credit usage. Satisfaction analytics track user feedback, such as satisfaction ratings and comments, which directly supports the CTO's need to monitor poor feedback on response quality. Tool use would help analyze whether tools are being invoked and succeeding, but it does not directly answer the abandoned-versus-resolved concern. Use is broader activity volume and does not target the response-quality feedback issue as directly as Satisfaction.

• Question 39:

  A company has an AI business solution that uses Microsoft Copilot Studio agents.

  You need to recommend prompt best practices to improve the effectiveness of agent interactions.

  Which two actions should you include in the recommendation? Each correct answer presents part of the solution.

  NOTE: Each correct selection is worth one point.

  A. Track the duration of the average user session.
  B. Analyze the prompt length distribution.
  C. Regularly test and refine the prompts based on user input.
  D. Use clear and specific instructions in the prompts.
  E. Measure system resource usage during prompt processing.
  C. Regularly test and refine the prompts based on user input.
  D. Use clear and specific instructions in the prompts.

  ---

  Explanation

  Testing and refining prompts based on user input plus using clear, specific instructions are core prompt-engineering practices. Clear instructions reduce ambiguity, and repeated testing exposes where the prompt fails with real user language.

  Measuring prompt length or system resource usage may be interesting, but those metrics do not directly improve interaction quality. Prompt quality improves through specificity, examples where useful, and iterative validation.

• Question 40:

  You need to design an application lifecycle management (ALM) process for a Microsoft Power Platform environment that contains a solution named Solution1.

  Solution1 must include a custom connector for Copilot in Microsoft Dynamics 365 Customer Service. Solution1 must meet the following requirements:

  1. Ensure that the custom connector can be deployed consistently across environments as part of the ALM process.

  2. Allow the custom connector to be edited only in the development environment.

  What should you include in the design?

  A. Add the custom connector to GitHub.
  B. Share the custom connector.
  C. Create the custom connector in the default solution.
  D. Add the custom connector to Solution1.
  D. Add the custom connector to Solution1.

  ---

  Explanation

  Adding the custom connector to the solution makes it part of the ALM package for consistent deployment across environments. When the solution is deployed as managed to production, the connector is controlled as a solution component instead of being edited directly in the target environment. Sharing the connector or keeping it only in the default solution would not provide repeatable deployment governance. Custom connectors should travel with the solution that depends on them.