Microsoft Azure AI Fundamentals (AI-901) practice questions

Domain coverage 2 domains

• Identify AI concepts and capabilities
• Implement AI solutions by using Microsoft Foundry

Sample questions 11 of 105 shown

1. Q1 Describe AI Workloads and Considerations (20–25%)

   An AI system used for loan application approvals rejects applicants from certain postal code areas at significantly higher rates than others, even when financial indicators are comparable. Which responsible AI principle is most directly violated?

   1. Privacy and security
   2. Fairness
   3. Transparency
   4. Accountability

   Fairness requires that AI systems treat all people fairly and distribute opportunities, resources, and information without bias. Disparate rejection rates correlated with postal codes, without equivalent financial justification, is a classic fairness failure mode.
2. Q2 Describe AI Workloads and Considerations (20–25%)

   A mortgage lender deploys an AI model to automate loan application decisions. The model achieves the required overall accuracy, does not expose applicant data, and shows users a brief explanation of the factors used. Testing reveals that equally qualified applicants from one age group receive significantly fewer approvals than others. Which responsible AI concern is most appropriate?

   1. Reliability and safety
   2. Privacy and security
   3. Transparency
   4. Fairness

   Fairness concerns whether AI systems treat individuals and groups equitably, especially when decisions affect opportunities like lending or hiring. The pattern reveals potential bias or disparate impact, which is a fairness issue.
3. Q3 Describe AI Workloads and Considerations (20–25%)

   A clinic plans to use AI capabilities to triage incoming patient messages. The clinic needs to use AI output to support triage, explain triage decisions when challenged, and ensure accountability if a patient is harmed by incorrect advice. Which approach is most appropriate?

   1. Rely on model confidence scores as the only supervision mechanism
   2. Assign a clinical lead to review and own AI-supported triage decisions
   3. Declare that the deployed model is accountable for triage outcomes
   4. Require patients to accept AI recommendations as final

   Accountability requires that individuals and organizations take responsibility for decisions made with AI support. AI can generate recommendations, but responsibility lies with the organization deploying it. A clinical lead should review and own triage decisions.
4. Q4 Describe AI Workloads and Considerations (20–25%)

   Which Microsoft service enables you to create custom question-answering solutions from FAQ documents, product manuals, or support content without writing code?

   1. Custom Question Answering in Azure AI Language
   2. Azure AI Bot Service
   3. Language Understanding (LUIS)
   4. Azure Cognitive Search

   Custom Question Answering (part of Azure AI Language) enables creating conversational knowledge bases from FAQ pages, product manuals, or other content without requiring coding.
5. Q5 Describe AI Workloads and Considerations (20–25%)

   A bank wants to use generative AI to draft customer hardship case suggestions for service agents. Output may affect payment plans, must be documented in case files, and must follow bank policy. Which oversight response is most appropriate?

   1. Use a larger model to reduce the need for review
   2. Only store the final AI recommendation
   3. Require agent review and approval before updating the case
   4. Let the model update the case when its confidence is high

   When AI outputs affect meaningful decisions, accountability and human oversight are critical. A human agent must review, verify, and approve AI suggestions before they affect customer cases, ensuring policy compliance and proper judgment.
6. Q6 Describe AI Workloads and Considerations (20–25%)

   A city services team wants to analyze resident feedback messages. The solution must find references to specific parks, departments, dates, phone numbers, and amounts so staff can route and summarize issues. Which text analysis capability is the best choice?

   1. Language detection
   2. Sentiment analysis
   3. Key phrase extraction
   4. Entity detection

   Entity detection identifies and labels specific references in unstructured text, such as places, organizations, dates, phone numbers, and quantities. This allows the system to locate structured references for routing and summarization.
7. Q7 Describe AI Workloads and Considerations (20–25%)

   A financial team receives scanned supplier invoices and PDF purchase orders. They need to capture vendor name, invoice number, total amount, and line item relationships. Which AI capability is the best choice?

   1. Image generation based on prompts
   2. Document and form information extraction
   3. Speech recognition and transcription
   4. General text sentiment analysis

   When the goal is to convert content from forms, invoices, or receipts into structured data, document and form information extraction is the best choice. It captures named fields, values, and relationships.
8. Q8 Describe AI Workloads and Considerations (20–25%)

   A team wants users to ask questions about product defects by typing a description and attaching a photo. Which model capability is the best choice?

   1. Multimodal model
   2. Speech recognition model
   3. Text-only generation model
   4. Image generation model

   Multimodal models handle multiple input types — text plus images, audio, or other modalities. In this scenario, the prompt includes both text and an attached photo, requiring a multimodal model.
9. Q9 Describe AI Workloads and Considerations (20–25%)

   A customer support team receives short ticket messages and wants to automatically identify specific items mentioned: product names, order numbers, and locations. Which text analysis technique best meets this need?

   1. Named entity recognition
   2. Language detection
   3. Key phrase extraction
   4. Sentiment analysis

   Named entity recognition finds and classifies specific entities in text. The support team needs structured details like product names, order numbers, and locations for ticket routing.
10. Q10 Describe AI Workloads and Considerations (20–25%)

    A marketing team wants an AI capability that drafts new product descriptions based on short prompts. Descriptions should naturally vary by audience and tone. Which approach is the best choice?

    1. Use an object detection model to identify product images
    2. Use a generative AI model to create text from prompts
    3. Use a fixed template library with keyword replacement
    4. Use a sentiment analysis model to score the prompt

    Generative AI creates new content — text, images, or audio — based on prompts by learning patterns from training data. This provides the flexibility to vary descriptions by audience and tone.
11. Q11 Describe AI Workloads and Considerations (20–25%)

    A marketing team wants to create a new visual asset for a product launch. The application must generate original banner images from text descriptions. Which model capability is the best choice?

    1. Speech recognition with transcription
    2. Image generation from text prompts
    3. Content Understanding for forms
    4. Computer vision image classification

    Image generation uses text prompts to create original visual assets. Computer vision analyzes existing images, while Content Understanding extracts information from content.

Answer all 105 questions with instant feedback →

Reader reviews 166 ratings

• My boss asked me to get the AI-901 cert for work. This was the best study tool I found. Passed in three weeks.

  ★★★★★★ Austin P.

• Straightforward and effective. No fluff in the AI-901 practice set, just relevant questions with solid answer keys.

  ★★★★★★ Owen P.

• Used this AI-901 prep extensively for three weeks. The progress tracking feature kept me accountable.

  ★★★★★★ Violet W.

• Really well-structured AI-901 practice set. I like that you can attempt questions multiple times and it tracks your progress.

  ★★★★★★ Lucas W.

• I was really impressed by the quality of the AI-901 questions. No typos, no vague wording — they clearly know the material.

  ★★★★★★ Stella P.

• Between the AI-901 practice questions and some hands-on labs, I felt fully prepared walking into the exam center.

  ★★★★★★ Caleb B.

Frequently asked questions

Free study resources

Community-verified analysis of 28 questions from real test-taker discussions — 7 deep analyses and 0 FAQs.

More Microsoft exams