AI-900 Practice Questions Part 1: 25 Microsoft Azure AI Fundamentals Questions

🕑 9 min read

Introduction

Artificial Intelligence is becoming a core part of modern applications, and Microsoft Azure provides services for machine learning, computer vision, natural language processing, document intelligence, and conversational AI.

In Part 1, we’ll go through the 25 practice questions, covering topics such as:

  • Machine Learning
  • Classification and Regression
  • Responsible AI
  • Computer Vision
  • Natural Language Processing
  • Anomaly Detection
  • Conversational AI
  • Feature Engineering

Each question includes the correct answer and a short explanation to make revision easier.

AI-900 Practice Questions Infographic

Question 1

A company employs customer service agents who provide telephone and email support.

The company develops a webchat bot to automatically answer common customer questions.

Which business benefit should the company expect?

A. Increased sales
B. Reduced workload for customer service agents
C. Improved product reliability
D. Increased hardware capacity

Answer: B. Reduced workload for customer service agents

Explanation: A chatbot can handle common questions automatically, reducing the workload of human agents.

Question 2

For a machine learning project, how should you split data for training and evaluation?

A. Use features for training and labels for evaluation
B. Randomly split the data into rows for training and rows for evaluation
C. Use labels for training and features for evaluation
D. Randomly split the data into columns

Answer: B. Randomly split the data into rows for training and evaluation

Explanation: Training and evaluation datasets are created by splitting observations (rows), not features (columns).

Question 3

A classification model produces the following confusion-matrix results:

  • True positives: 11
  • False positives: 5
  • False negatives: 1,033
  • True negatives: 13,951

How many correctly predicted positives and false negatives are there?

A. 5 and 11
B. 11 and 1,033
C. 1,033 and 13,951
D. 13,951 and 5

Answer: B. 11 and 1,033

Explanation: Correctly predicted positives are true positives (11), while incorrectly predicted negatives are false negatives (1,033).

Question 4

You build a machine learning model using the automated machine learning interface.

You need the model to meet Microsoft’s transparency principle for responsible AI.

What should you do?

A. Set Validation type to Auto
B. Enable Explain best model
C. Set Primary metric to accuracy
D. Set Max concurrent iterations to 0

Answer: B. Enable Explain best model

Explanation: Model explainability helps users understand how features influence predictions, supporting transparency.

Question 5

Which statements correctly describe anomaly detection?

Statement 1: Forecasting housing prices from historical data is anomaly detection.
Statement 2: Identifying suspicious sign-ins based on deviations from normal patterns is anomaly detection.
Statement 3: Predicting whether a patient will develop diabetes is anomaly detection.

A. Yes, Yes, Yes
B. No, Yes, No
C. Yes, No, Yes
D. No, No, Yes

Answer: B. No, Yes, No

Explanation: Housing-price forecasting is regression, suspicious-pattern detection is anomaly detection, and diabetes prediction is classification.

Question 6

Handling unusual or missing values provided to an AI system is primarily a consideration for which Microsoft responsible AI principle?

A. Inclusiveness
B. Privacy and security
C. Reliability and safety
D. Transparency

Answer: C. Reliability and safety

Explanation: AI systems should behave reliably and safely under both expected and unexpected conditions.

Question 7

Match each scenario with the appropriate AI workload.

  1. An automated chat that answers questions about refunds and exchanges
  2. Determining whether a photo contains a person
  3. Determining whether a review is positive or negative

A. Conversational AI, Computer Vision, Natural Language Processing
B. NLP, Anomaly Detection, Computer Vision
C. Computer Vision, NLP, Conversational AI
D. Anomaly Detection, Computer Vision, Knowledge Mining

Answer: A

Explanation: Chatbots use conversational AI, image analysis uses computer vision, and sentiment analysis uses NLP.

Question 8

You are designing an AI system that empowers people with hearing, visual, and other impairments.

Which responsible AI principle does this represent?

A. Fairness
B. Inclusiveness
C. Reliability and safety
D. Accountability

Answer: B. Inclusiveness

Explanation: Inclusiveness ensures AI technology considers and supports people with diverse needs and abilities.

Question 9

Match the requirement with the responsible AI principle:

  1. AI operates as designed and responds safely to unexpected conditions
  2. Humans oversee decisions made by AI systems
  3. Consumers have information and control over how their data is used

A. Reliability and safety → Accountability → Privacy and security
B. Fairness → Transparency → Inclusiveness
C. Accountability → Fairness → Reliability and safety
D. Transparency → Inclusiveness → Accountability

Answer: A

Explanation: These requirements correspond respectively to reliability and safety, accountability, and privacy and security.

Question 10

When developing an AI system for self-driving cars, which responsible AI principle should ensure consistent operation during unexpected circumstances?

A. Inclusiveness
B. Accountability
C. Reliability and safety
D. Fairness

Answer: C. Reliability and safety

Explanation: Reliability and safety require AI systems to respond safely to unexpected operating conditions.

Question 11

You are building an AI system.

Which task helps meet Microsoft’s transparency principle?

A. Ensure all visuals have text readable by screen readers
B. Enable autoscaling
C. Provide documentation to help developers debug code
D. Ensure the training dataset represents the population

Answer: C. Provide documentation to help developers debug code

Explanation: Documentation helps people understand how the AI system works, supporting transparency.

Question 12

Match the scenario with the appropriate AI workload.

  1. Identify handwritten letters
  2. Predict the sentiment of a social media post
  3. Identify a fraudulent credit card payment
  4. Predict next month’s toy sales

A. Computer Vision, NLP, Anomaly Detection, Regression
B. NLP, Computer Vision, Regression, Anomaly Detection
C. Computer Vision, Regression, NLP, Classification
D. Anomaly Detection, NLP, Computer Vision, Regression

Answer: A

Explanation: Image recognition uses computer vision, sentiment uses NLP, fraud detection uses anomaly detection, and numeric forecasting uses regression.

Question 13

A company is exploring voice-recognition technologies for smart-home devices and wants to identify barriers that might unintentionally exclude specific groups.

Which responsible AI principle does this represent?

A. Accountability
B. Fairness
C. Inclusiveness
D. Privacy and security

Answer: C. Inclusiveness

Explanation: Inclusiveness aims to ensure AI technologies accommodate a broad range of human needs and experiences.

Question 14

Which three are Microsoft guiding principles for responsible AI?

A. Knowledgeability
B. Decisiveness
C. Inclusiveness
D. Fairness
E. Opinionatedness
F. Reliability and safety

Answer: C, D and F

Explanation: Inclusiveness, fairness, and reliability and safety are among Microsoft’s six responsible AI principles.

Question 15

Returning a bounding box that identifies the location of a vehicle in an image is an example of:

A. Image classification
B. Object detection
C. Optical Character Recognition
D. Semantic segmentation

Answer: B. Object detection

Explanation: Object detection identifies objects and returns their locations, commonly using bounding boxes.

Question 16

Which machine learning process is used to generate additional features?

A. Feature engineering
B. Feature selection
C. Model evaluation
D. Model training

Answer: A. Feature engineering

Explanation: Feature engineering transforms existing data or creates additional features that can improve model learning.

Question 17

You need to analyze photos and ensure they:

  • Contain at least one face
  • Contain at least one person wearing sunglasses

What should you use?

A. Verify operation in the Face service
B. Detect operation in the Face service
C. Describe Image operation
D. Analyze Image operation

Answer: B. Detect operation in the Face service

Explanation: According to the supplied material, face detection can detect faces and the glasses attribute.

Question 18

When designing an AI system to determine whether loans should be approved, the factors used to make decisions should be explainable.

Which responsible AI principle does this demonstrate?

A. Transparency
B. Inclusiveness
C. Fairness
D. Privacy and security

Answer: A. Transparency

Explanation: Transparency helps people understand the data, algorithms, and factors involved in AI-generated decisions.

Question 19

For each statement, determine whether it correctly represents the named responsible AI principle.

  1. Explaining the outcome of a credit-loan application demonstrates transparency.
  2. Prioritizing insurance claims based on injuries demonstrates reliability and safety.
  3. Offering an AI solution at different prices in different sales territories demonstrates inclusiveness.

A. Yes, Yes, Yes
B. Yes, No, No
C. No, Yes, No
D. No, No, Yes

Answer: B. Yes, No, No

Explanation: Explaining AI decisions supports transparency; the other two scenarios do not demonstrate the stated principles.

Question 20

Match the responsible AI principle with its requirement.

  1. The system must not discriminate based on gender or race
  2. Personal data must be visible only to approved users
  3. Automated decisions must be recorded so approved users can identify why a decision was made

A. Fairness, Privacy and security, Transparency
B. Transparency, Fairness, Accountability
C. Inclusiveness, Transparency, Privacy and security
D. Privacy and security, Reliability and safety, Fairness

Answer: A

Explanation: Non-discrimination relates to fairness, data protection to privacy and security, and explainable decision records to transparency.

Question 21

You plan to deploy an Azure Machine Learning model as a service.

Which three processes should you perform in sequence before deployment?

A. Model training → Data preparation → Model evaluation
B. Data preparation → Model training → Model evaluation
C. Model evaluation → Data preparation → Model training
D. Data encryption → Model evaluation → Model retraining

Answer: B. Data preparation → Model training → Model evaluation

Explanation: Data is prepared first, the model is then trained, and finally its performance is evaluated before deployment.

Question 22

You are building an AI-based application and need to follow responsible AI principles.

Which two actions should you take?

A. Implement an Agile software development methodology
B. Implement AI model validation as part of the software review process
C. Establish a risk governance committee including legal, risk, and privacy representatives
D. Prevent disclosure of AI-based automated decision-making

Answer: B and C

Explanation: Model validation supports reliability and safety, while governance involving legal, risk, and privacy stakeholders supports responsible AI practices.

Question 23

According to Microsoft’s responsible AI principles, AI systems should not reflect biases from the datasets used to train them.

Which principle does this represent?

A. Accountability
B. Fairness
C. Inclusiveness
D. Transparency

Answer: B. Fairness

Explanation: Fairness requires AI systems to avoid discriminatory outcomes or biases against individuals or groups.

Question 24

An AI system produces different outcomes for people solely because of characteristics such as gender or race.

Which responsible AI principle is most directly violated?

A. Accountability
B. Fairness
C. Inclusiveness
D. Transparency

Answer: B. Fairness

Explanation: Fairness requires AI systems to avoid discriminatory decisions based on characteristics such as gender or race.

Question 25

Match the AI workload to the appropriate scenario.

  1. An automated chatbot answers questions about refunds and exchanges
  2. Determine whether a photo contains a person
  3. Determine whether a review is positive or negative

A. Knowledge Mining, Computer Vision, Natural Language Processing
B. Computer Vision, Knowledge Mining, Anomaly Detection
C. NLP, Computer Vision, Regression
D. Anomaly Detection, NLP, Knowledge Mining

Answer: A

Explanation: The supplied material maps these scenarios to knowledge mining, computer vision, and natural language processing respectively.

Conclusion

That completes Part 1 of the AI-900 Practice Questions series — Questions 1–25.

This set covered several important AI-900 concepts, including machine learning fundamentals, responsible AI, computer vision, natural language processing, anomaly detection, conversational AI, feature engineering, and model evaluation.

One of the most important areas to revise from this set is Microsoft’s Responsible AI principles:

Fairness • Reliability and Safety • Privacy and Security • Inclusiveness • Transparency • Accountability

Also remember the basic differences between common AI workloads:

Regression → Predict a numerical value
Classification → Predict a category
Anomaly Detection → Identify unusual behavior
Computer Vision → Analyze images and visual information
NLP → Understand and analyze language
Conversational AI → Enable human-like interactions

Keep practicing these concepts rather than simply memorizing answers—the AI-900 exam tests whether you can identify the correct AI workload or Azure AI capability from a scenario.

Connect with Me on LinkedIn

Thank you for reading! If you found this article helpful, connect with me on LinkedIn for more learning resources, technology insights, certification content, and career tips. Let’s learn and grow together.