5.3. Practice Questions
Domain 1: Business Value of Generative AI
Question 1
A financial services company wants to automatically categorize incoming customer emails as "complaint," "inquiry," or "feedback" for routing to appropriate teams. What type of AI is most appropriate?
A. Generative AI using Microsoft 365 Copilot
B. Traditional AI/ML classification model
C. Microsoft Foundry with GPT-4
D. Copilot Studio custom agent
Answer: B
Rationale: This is a classification task—assigning categories to inputs based on patterns. Classification is a traditional AI/ML use case, not generative AI. Generative AI creates content; classification AI categorizes content. Options A, C, and D all use generative AI, which is overkill and less appropriate for pure classification.
Question 2
An employee complains that Copilot's responses about company vacation policy are sometimes incorrect. The IT team suggests fine-tuning the model on company documents. What is the BEST recommendation?
A. Fine-tune the model as suggested
B. Ground Copilot on the correct policy documents using RAG
C. Switch to a more advanced AI model
D. Accept that AI will sometimes be wrong
Answer: B
Rationale: The issue is that Copilot lacks access to authoritative company documents, not that it needs to behave differently. Grounding/RAG provides access to specific documents. Fine-tuning changes model behavior (how it responds), not what documents it can access. This is a common distinction the exam tests.
Question 3
Which metric BEST indicates business value from a Microsoft 365 Copilot deployment?
A. Number of Copilot licenses deployed
B. Percentage of employees who have activated Copilot
C. Time saved on tasks where Copilot is used
D. Number of Copilot queries submitted per day
Answer: C
Rationale: Business value = outcomes, not activity. Time saved directly measures productivity improvement. Licenses deployed (A) measures investment, not value. Activation (B) measures setup, not usage. Queries (D) measures activity, not value—users could submit many queries with no productivity gain.
Question 4
A marketing team uses AI to generate customer-facing content without human review. What is the PRIMARY risk?
A. Higher costs from token usage
B. AI fabricating incorrect information that damages credibility
C. Slower content production
D. Employees becoming dependent on AI
Answer: B
Rationale: Fabrication (hallucination) is the primary risk of unreviewed AI-generated content. AI can generate confident but incorrect statements that damage credibility when published to customers. Token costs (A) are operational, not risk. AI typically speeds production (C), not slows it. Dependency (D) is a concern but secondary to factual accuracy.
Question 5
What is the PRIMARY purpose of prompt engineering?
A. To reduce the cost of AI inference
B. To improve the quality and relevance of AI responses
C. To train AI models on specific data
D. To connect AI to external databases
Answer: B
Rationale: Prompt engineering is the skill of crafting effective prompts to get better outputs. It directly improves quality and relevance. While good prompts can indirectly reduce costs by avoiding iterations, that's not the primary purpose. Training models (C) is fine-tuning. Connecting to databases (D) is grounding/RAG.
Domain 2: Microsoft AI Apps and Services
Question 6
A company uses the free Microsoft Copilot and finds it cannot answer questions about their recent sales meeting. Why?
A. The free version has slower response times
B. The free version lacks access to organizational data through Microsoft Graph
C. The free version uses an older AI model
D. The free version requires additional configuration
Answer: B
Rationale: The critical difference between free Microsoft Copilot and Microsoft 365 Copilot is organizational data access via Microsoft Graph. Free Copilot cannot access your emails, calendar, documents, or meetings. This is one of the most frequently tested distinctions on the exam.
Question 7
An HR department wants a chatbot that answers employee questions about benefits, grounded on their benefits documentation. The IT team proposes building it in Microsoft Foundry. What is the BEST recommendation?
A. Proceed with Foundry for maximum flexibility
B. Use Microsoft 365 Copilot instead
C. Use Copilot Studio to build a custom agent
D. Use Azure AI Search directly
Answer: C
Rationale: Copilot Studio is designed for exactly this use case—custom agents grounded on specific knowledge bases. Foundry is overkill for a business chatbot (it's for building AI products and custom model development). M365 Copilot is for standard productivity, not custom FAQ bots. Azure AI Search is a component, not a complete solution.
Question 8
A finance manager needs to analyze three years of expense data to identify anomalies and statistical trends. Standard Excel formulas are insufficient. What would you recommend?
A. Microsoft 365 Copilot in Excel
B. Researcher agent
C. Analyst agent
D. Microsoft Foundry
Answer: C
Rationale: Analyst agent is designed for advanced data analysis using chain-of-thought reasoning and Python code execution. It can handle complex statistical analysis that standard Excel Copilot cannot. Researcher is for multi-source research synthesis, not data analysis. Foundry is overkill for analysis tasks.
Question 9
A development team needs to add automatic invoice data extraction to a custom application. What should they use?
A. Microsoft 365 Copilot
B. Copilot Studio
C. Azure AI Document Intelligence
D. Microsoft Foundry Agent Service
Answer: C
Rationale: Azure AI Document Intelligence is specifically designed for extracting structured data from documents like invoices—a developer service for embedding in custom applications. M365 Copilot is for end-user productivity, not custom application development. Copilot Studio is for building agents, not document processing.
Question 10
Which capability differentiates Microsoft 365 Copilot from the free Microsoft Copilot?
A. Ability to search the web
B. Access to GPT-4 model
C. Organizational data access through Microsoft Graph
D. Ability to generate images
Answer: C
Rationale: Organizational data access through Microsoft Graph is THE key differentiator. This enables M365 Copilot to answer questions about your meetings, documents, emails, and colleagues. Both versions can search the web, use advanced models, and generate content—but only M365 Copilot knows about YOUR organization.
Domain 3: Implementation and Adoption Strategy
Question 11
An AI loan approval system shows different approval rates for different demographic groups with similar credit profiles. Which responsible AI principle is MOST relevant?
A. Transparency
B. Fairness
C. Reliability & Safety
D. Inclusiveness
Answer: B
Rationale: Fairness means AI should treat all people equitably. Different outcomes for similar profiles across demographic groups indicates unfair treatment. Transparency is about explaining decisions. Reliability is about consistent performance. Inclusiveness is about empowering everyone, including accessibility.
Question 12
Six months after deploying Microsoft 365 Copilot, analytics show most employees haven't tried it, but those who have tried it continue using it. Users report they "don't have time to learn new tools." What type of adoption barrier is this?
A. Capability barrier
B. Motivation barrier
C. Opportunity barrier
D. Technology barrier
Answer: C
Rationale: "Don't have time" indicates an Opportunity barrier—users lack the time/chance to engage with the tool. It's not Capability (they don't say they can't learn) or Motivation (those who try it keep using it, showing value). The intervention needed is protected learning time, not training or success stories.
Question 13
What is the PRIMARY purpose of an AI council?
A. To develop and deploy AI solutions
B. To provide AI training to employees
C. To set AI strategy, standards, and governance
D. To manage AI licenses and budgets
Answer: C
Rationale: AI councils focus on strategy, standards, and governance—the strategic/oversight layer. Developing solutions (A) is operational. Training (B) is HR/L&D function. License management (D) is operational. Councils decide WHAT AI initiatives to pursue and HOW to govern them, not day-to-day execution.
Question 14
A CFO wants predictable monthly AI costs. Which licensing model provides the MOST predictability?
A. Azure AI Services consumption-based pricing
B. Microsoft 365 Copilot per-user subscription
C. Microsoft Foundry compute-based pricing
D. Copilot Studio message-pack pricing
Answer: B
Rationale: Per-user subscription provides highest predictability—fixed cost per user per month. Consumption-based (A) and compute-based (C) pricing varies with usage. Message-pack pricing (D) has some predictability but varies with agent usage volume.
Question 15
Which of the following is NOT one of Microsoft's six responsible AI principles?
A. Fairness
B. Profitability
C. Transparency
D. Accountability
Answer: B
Rationale: Microsoft's six responsible AI principles are: Fairness, Reliability & Safety, Privacy & Security, Inclusiveness, Transparency, and Accountability. Profitability is a business objective, not a responsible AI principle.
Mixed Domain Questions
Question 16
A healthcare organization wants AI to help triage patient inquiries by recommending urgency levels. What are the TWO most important considerations?
A. Using the newest AI model for accuracy
B. Ensuring fairness across patient demographics
C. Minimizing token costs
D. Implementing human oversight for safety
E. Maximizing automation to reduce staff workload
Answer: B and D
Rationale: Healthcare AI affecting patient care requires fairness (equitable treatment across demographics) and reliability/safety (human oversight for medical decisions). Model newness (A) doesn't ensure appropriateness. Cost (C) and efficiency (E) are secondary to safety and equity in healthcare.
Question 17
A company successfully piloted Microsoft 365 Copilot with 50 users and wants to roll out to 5,000 employees. What should they do FIRST?
A. Purchase licenses for all 5,000 users
B. Develop a comprehensive training program
C. Establish an AI champions program
D. Create governance policies and an adoption plan
Answer: D
Rationale: Before scaling, establish governance and adoption strategy. Without governance, you risk misuse. Without an adoption plan, you risk low usage. Training (B) and champions (C) are part of the adoption plan but come after strategic planning. Purchasing licenses (A) before planning wastes investment if adoption fails.
Question 18
Which scenario BEST justifies using Microsoft Foundry instead of Copilot Studio?
A. Building a custom HR chatbot
B. Creating an IT helpdesk agent
C. Building a proprietary AI product to sell to customers
D. Extending Microsoft 365 Copilot with custom actions
Answer: C
Rationale: Foundry is for building AI products—proprietary solutions that may be sold or differentiate the business. HR chatbots (A), IT helpdesk (B), and Copilot extensions (D) are Copilot Studio use cases. Foundry's complexity and cost are justified only for product development or specialized model requirements.
Question 19
What is the PRIMARY benefit of grounding Microsoft 365 Copilot on organizational data through Microsoft Graph?
A. Reduced token costs
B. Faster response times
C. Relevant, context-aware responses about your organization
D. Elimination of the need for user training
Answer: C
Rationale: Grounding enables Copilot to provide relevant responses about YOUR organization—your meetings, documents, colleagues, and context. Without grounding, responses are generic. Grounding doesn't directly reduce costs (A) or speed (B), and users still need training (D).
Question 20
An employee advocates strongly for AI adoption within their team, demonstrates Copilot use cases in meetings, and helps colleagues with prompts. This person's role is BEST described as:
A. AI council member
B. AI champion
C. IT support specialist
D. Change management consultant
Answer: B
Rationale: This describes an AI champion—a peer advocate who demonstrates value and provides informal support within their team. AI council members (A) focus on strategy and governance, not peer-level advocacy. IT support (C) handles technical issues. Change management consultants (D) are formal project roles.
