Microsoft AB-731 Online Practice Questions and Exam Preparation

Microsoft AB-731 Online Questions & Answers

• Question 1:

  DRAG DROP

  Match the Microsoft responsible AI principles to the appropriate descriptions. To answer, drag the principle from the column on the left to its description on the right. Each principle may be used once, more than once, or not at all.

  NOTE: Each correct match is worth one point.

  Select and Place:

  

  

  ---

  The answer image maps privacy and security to protecting personal information, transparency to explaining how and why decisions are made, inclusiveness to accessible design, and accountability to human oversight and control of AI systems.

• Question 2:

  Your company uses a fine-tuned generative AI solution trained on data that is representative of the general population.

  You discover that some of the generated responses include inappropriate or exclusionary language based on ableist assumptions.

  You need to prevent the inappropriate responses. Your solution must minimize costs.

  What should you do?

  A. Apply a newer version of the generative AI model.
  B. Apply a content-moderation filter.
  C. Create a new version of the solution that is trained on only inclusive and representative content.
  D. Create a new version of the solution that is trained on only exclusionary content.
  B. Apply a content-moderation filter.

  ---

  Explanation

  A content-moderation filter is the lower-cost mitigation for inappropriate or exclusionary generated language. Retraining or replacing the model can be more expensive, and training on exclusionary content would worsen the problem.

• Question 3:

  HOTSPOT

  An organization is implementing trustworthy AI practices. The organization wants to create a governance structure that provides oversight, supports responsible AI adoption, and aligns AI initiatives across departments.

  Select the answer that correctly completes the sentence.

  

  

  ---

  The answer image selects guiding AI strategy, ensuring responsible AI oversight, and promoting alignment across business units. That is the governance role of an AI council, unlike deployment configuration or day-to-day labeling work.

• Question 4:

  A business team reports that a generative AI solution gives answers that are too broad and inconsistent. The team wants to improve the prompt before changing the model.

  Which two prompt improvements should you recommend?

  A. Include the task goal, relevant context, output format, and success criteria
  B. Provide examples that show the expected response style and level of detail
  C. Use only product acronyms to keep the prompt short
  D. Ask the model to ignore source documents when generating the answer
  A. Include the task goal, relevant context, output format, and success criteria
  B. Provide examples that show the expected response style and level of detail

  ---

  Explanation

  Clear prompts should state the task goal, context, output format, and success criteria, and examples help show the expected style and detail. Acronym-only prompts and ignoring source documents make responses less grounded and less consistent.

• Question 5:

  HOTSPOT

  A company is estimating the cost of using generative AI language models. The company wants to understand the common pricing unit used to measure model input and output processing.

  Select the answer that correctly completes the sentence.

  

  

  ---

  The answer image selects tokens. Generative AI language model cost is typically measured by the number of input and output tokens processed, not by files, requests alone, or words alone.

• Question 6:

  Your company uses a generative AI solution.

  You need to improve the quality of responses by using grounding.

  Which statement accurately describes how grounding improves accuracy and relevancy?

  A. references a diverse set of people, disciplines, and perspectives
  B. explains how and why AI models generate content
  C. anchors the responses in specific data sources
  D. specifies the strengths and weaknesses of the AI model
  C. anchors the responses in specific data sources

  ---

  Explanation

  Grounding improves accuracy by anchoring generated responses in specific trusted data sources. It does not explain model internals or list model strengths; it supplies relevant context that helps the model produce more useful and verifiable output.

• Question 7:

  HOTSPOT

  You are helping business users understand the types of productivity tasks that Microsoft 365 Copilot can perform in Microsoft 365 apps.

  Select the answer that correctly completes the sentence.

  

  

  ---

  The answer image should be No, Yes, and Yes. Vision tools are not for extracting key phrases from PDFs, but they can generate image labels and natural language descriptions, while Document Intelligence can process invoices and credit notes.

• Question 8:

  Your company wants to ensure that AI solutions are used responsibly and align with company values and compliance requirements.

  You need to establish governance principles for AI use.

  Which two actions should you perform?

  A. Allow each department to tailor governance processes for its own AI initiatives.
  B. Define accountability norms for AI decisions across business and technical teams.
  C. Create a process to review AI initiatives for responsible AI alignment.
  D. Assign governance ownership primarily to the AI engineering and data science teams.
  E. Focus governance efforts on AI systems that handle regulated or sensitive data.
  B. Define accountability norms for AI decisions across business and technical teams.
  C. Create a process to review AI initiatives for responsible AI alignment.

  ---

  Explanation

  Responsible governance needs shared accountability norms and a formal review process for AI initiatives. Limiting ownership to technical teams, fragmenting governance by department, or reviewing only sensitive systems leaves business and compliance risks unmanaged.

• Question 9:

  HOTSPOT

  A company needs to analyze internal documents before sharing generated responses with users. The solution must detect Personally Identifiable Information (PII) and redact the PII from the output by using a prebuilt Azure AI capability.

  Select the answer that correctly completes the sentence.

  

  

  ---

  The answer image selects Azure Language in Foundry Tools. PII detection and redaction for text is a language capability; search, vision, and document extraction are not the direct text PII redaction service.

• Question 10:

  Your company plans to use generative AI to help build a website that will showcase various existing products.

  Which capability best describes a benefit of using generative AI for this project?

  A. creating product descriptions based on product specifications
  B. designing a new product based on customer requirements
  C. translating product descriptions into a different language
  D. analyzing product sales to identify patterns and trends
  A. creating product descriptions based on product specifications

  ---

  Explanation

  Creating product descriptions from product specifications is a direct generative AI content-creation use case. Translation changes language, sales analysis finds patterns, and new product design is a different business activity from writing website copy.