AI-900 sample exam questions

15 exam-style questions for Microsoft AI-900: Azure AI Fundamentals, with answers and explanations. Take them as a timed practice exam first if you want a score out of 1000.

Take the free practice exam AI-900 exam guide
  1. Question 1Single choiceAI workloads and responsible AI
    You are building an AI system.
    Which task should you include to ensure that the service meets the Microsoft transparency principle for responsible AI?
    1. A
      Ensure that all visuals have an associated text that can be read by a screen reader.
    2. B
      Enable autoscaling to ensure that a service scales based on demand.
    3. C
      Provide documentation to help developers debug code.
    4. D
      Ensure that a training dataset is representative of the population.
    Show answer

    Correct answer: C

    Why

    Transparency means people should understand how the system works; documentation that helps developers understand and debug the code supports that. Screen-reader text is inclusiveness, autoscaling is reliability, and a representative dataset is fairness.

  2. Question 2Single choiceComputer vision
    You run a charity event that involves posting photos of people wearing sunglasses on Twitter.
    You need to ensure that you only retweet photos that meet the following requirements:
    ✑ Include one or more faces.
    ✑ Contain at least one person wearing sunglasses.
    What should you use to analyze the images?
    1. A
      the Verify operation in the Face service
    2. B
      the Detect operation in the Face service
    3. C
      the Describe Image operation in the Computer Vision service
    4. D
      the Analyze Image operation in the Computer Vision service
    Show answer

    Correct answer: B

    Why

    The Face service Detect operation finds faces and can return attributes such as glasses type, including sunglasses, so one call checks both requirements. Verify compares two faces for identity, and Describe Image only produces a caption without reliable glasses detail.

  3. Question 3Single choiceAI workloads and responsible AI
    When you design an AI system to assess whether loans should be approved, the factors used to make the decision should be explainable.
    This is an example of which Microsoft guiding principle for responsible AI?
    1. A
      transparency
    2. B
      inclusiveness
    3. C
      fairness
    4. D
      privacy and security
    Show answer

    Correct answer: A

    Why

    Transparency requires that people can understand how an AI system reaches its decisions, so making loan-approval factors explainable is transparency. Fairness concerns avoiding bias between groups, not explaining the decision logic.

  4. Question 4Single choiceMachine learning
    Which metric can you use to evaluate a classification model?
    1. A
      true positive rate
    2. B
      mean absolute error (MAE)
    3. C
      coefficient of determination (R2)
    4. D
      root mean squared error (RMSE)
    Show answer

    Correct answer: A

    Why

    No community answer was recorded. True positive rate (recall) is derived from the confusion matrix and is a classification metric. MAE, R2 and RMSE all measure numeric error and are used to evaluate regression models.

  5. Question 5Multiple responseComputer vision
    In which two scenarios can you use the Form Recognizer service? Each correct answer presents a complete solution.
    NOTE: Each correct selection is worth one point.
    1. A
      Extract the invoice number from an invoice.
    2. B
      Translate a form from French to English.
    3. C
      Find image of product in a catalog.
    4. D
      Identify the retailer from a receipt.
    Show answer

    Correct answer: A, D

    Why

    Form Recognizer (now Azure AI Document Intelligence) extracts fields from documents, such as the invoice number from an invoice and the merchant name from a receipt. Translation needs Translator, and finding product images needs a vision service.

  6. Question 6Multiple responseNatural language processing
    You need to provide content for a business chatbot that will help answer simple user queries.
    What are three ways to create question and answer text by using QnA Maker? Each correct answer presents a complete solution.
    NOTE: Each correct selection is worth one point.
    1. A
      Generate the questions and answers from an existing webpage.
    2. B
      Use automated machine learning to train a model based on a file that contains the questions.
    3. C
      Manually enter the questions and answers.
    4. D
      Connect the bot to the Cortana channel and ask questions by using Cortana.
    5. E
      Import chit-chat content from a predefined data source.
    Show answer

    Correct answer: A, C, E

    Why

    QnA Maker knowledge bases can be populated by extracting Q&A pairs from existing webpages or files, by entering pairs manually, and by adding predefined chit-chat content. Automated ML and the Cortana channel do not create knowledge base content.

  7. Question 7Multiple responseAI workloads and responsible AI
    Which two scenarios are examples of a conversational AI workload? Each correct answer presents a complete solution.
    NOTE: Each correct selection is worth one point.
    1. A
      a smart device in the home that responds to questions such as ג€What will the weather be like today?ג€
    2. B
      a website that uses a knowledge base to interactively respond to usersג€™ questions
    3. C
      assembly line machinery that autonomously inserts headlamps into cars
    4. D
      monitoring the temperature of machinery to turn on a fan when the temperature reaches a specific threshold
    Show answer

    Correct answer: A, B

    Why

    No community answer was recorded. A smart home device answering spoken questions and a website that converses with users using a knowledge base are conversational AI. Autonomous assembly machinery and threshold-based fan control involve no dialog with users.

  8. Question 8Drag and dropNatural language processing
    DRAG DROP

    Match the services to the appropriate descriptions.

    To answer, drag the appropriate service from the column on the left to its description on the right. Each service 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.

    Move each item to the matching row:

    • Azure Storage
    • Azure Bot Service
    • Language Service
    • Speech
    • Enables the use of natural language to query a knowledge base.
    • Enables the real-time transcription of speech-to-text.
    Show answer
    • Enables the use of natural language to query a knowledge base. Language Service
    • Enables the real-time transcription of speech-to-text. Speech
    Why

    The Language service's custom question answering (successor to QnA Maker) lets users query a knowledge base in natural language. The Speech service provides real-time speech-to-text transcription. Bot Service only hosts the bot, and Azure Storage just stores data.

  9. Question 9Single choiceMachine learning
    Predicting how many vehicles will travel across a bridge on a give day is an example of _______.

    Select the answer that correctly completes the sentence.
    1. A
      regression
    2. B
      translation
    3. C
      classification
    4. D
      clustering
    Show answer

    Correct answer: A

    Why

    The number of vehicles crossing a bridge is a numeric value, so predicting it is regression. Classification predicts categories, clustering groups unlabeled data and translation converts language.

  10. Question 10HotspotMachine learning
    HOTSPOT

    Select the answer that correctly completes the sentence.

    A banking system that predicts whether a loan will be repaid is an example of the ___ type of machine learning.clustering / regression / classification
    Show answer
    • A banking system that predicts whether a loan will be repaid is an example of the ___ type of machine learning. classification
    Why

    Whether a loan will be repaid is a yes/no outcome, so predicting it is binary classification. Regression predicts numeric values and clustering groups unlabeled data.

  11. Question 11HotspotComputer vision
    HOTSPOT

    To complete the sentence, select the appropriate option in the answer area.

    An AI solution that helps photographers take better portrait photographs by providing feedback on exposure, noise, and occlusion is an example of facial ___analysis. / detection. / recognition.
    Show answer
    • An AI solution that helps photographers take better portrait photographs by providing feedback on exposure, noise, and occlusion is an example of facial ___ detection.
    Why

    Exposure, noise and occlusion are quality attributes returned by the Face service detect operation, so this is facial detection. Recognition identifies who a person is. Some study material files these attributes under facial analysis, but detection is the intended answer.

  12. Question 12Single choiceNatural language processing
    An app that analyzes social media posts to identify their tone is an example of which type of natural language processing (NLP) workload?
    1. A
      sentiment analysis
    2. B
      speech recognition
    3. C
      key phrase extraction
    4. D
      entity recognition
    Show answer

    Correct answer: A

    Why

    Identifying the tone of social media posts, such as positive, negative or neutral, is sentiment analysis. Speech recognition transcribes audio, key phrase extraction finds main topics and entity recognition finds named items.

  13. Question 13HotspotNatural language processing
    HOTSPOT

    For each of the following statements, select Yes if the statement is true. Otherwise, select No.

    NOTE: Each correct selection is worth one point.

    You can use Language Service's question answering to query an Azure SQL database.Yes / No
    You should use Language Service's question answering when you want a knowledge base to provide the same answer to different users who submit similar questions.Yes / No
    Language Service's question answering can determine the intent of a user utterance.Yes / No
    Show answer
    • You can use Language Service's question answering to query an Azure SQL database. No
    • You should use Language Service's question answering when you want a knowledge base to provide the same answer to different users who submit similar questions. Yes
    • Language Service's question answering can determine the intent of a user utterance. No
    Why

    Question answering returns answers from a knowledge base built from FAQs, documents or manual entries, giving the same answer to similar questions; it does not query databases such as Azure SQL. Determining intent is conversational language understanding, not question answering.

  14. Question 14Drag and dropAI workloads and responsible AI
    DRAG DROP

    You are designing a system that will generate insurance quotes automatically.

    Match the Microsoft responsible AI principles to the appropriate requirements.

    To answer, drag the appropriate principle from the column on the left to its requirement on the right. Each principle may be used once, more than once, or not at all.

    NOTE: Each correct match is worth one point.

    Move each item to the matching row:

    • Accountability
    • Fairness
    • Inclusiveness
    • Privacy and security
    • Reliability and safety
    • Transparency
    • A customer's personal information must be visible only to staff who are involved in the decision-making process:
    • The decision-making process must be recorded so that staff can identify the reasoning behind a particular quote:
    • The system must be accessible to customers who use screen readers or other assistive technology:
    Show answer
    • A customer's personal information must be visible only to staff who are involved in the decision-making process: Privacy and security
    • The decision-making process must be recorded so that staff can identify the reasoning behind a particular quote: Transparency
    • The system must be accessible to customers who use screen readers or other assistive technology: Inclusiveness
    Why

    Restricting who can see personal data is privacy and security. Recording the decision process so staff can understand why a quote was produced is transparency. Supporting screen readers and assistive technology is inclusiveness.

  15. Question 15Single choiceComputer vision
    You need to identify street names based on street signs in photographs.

    Which type of computer vision should you use?
    1. A
      object detection
    2. B
      optical character recognition (OCR)
    3. C
      image classification
    4. D
      facial recognition
    Show answer

    Correct answer: B

    Why

    Reading street names from signs means extracting text from images, which is OCR. Object detection could locate the sign but not read it, and image classification and facial recognition do not read text.