AI-901 Practice Test 3 – 50 Questions and Answers (Azure AI Fundamentals)

AI-901 Practice Test 3: 50 Questions and Answers for Azure AI Fundamentals

Published July 5, 2026

Practice AI-901 questions for the Microsoft Azure AI Fundamentals certification. This free quiz covers AI concepts, responsible AI, generative AI, Microsoft Foundry, AI agents, language, speech, vision, and content understanding. Select your answers, review the explanations, and track your progress in the Learning Dashboard.

Exam: AI-901Questions: 50Recommended score: 70%+Time: 60 minutes

Before you start

This AI-901 practice test includes single-choice, multiple-response, and true/false questions. When a question requires more than one answer, the question text tells you exactly how many answers to choose.

This is an independent educational practice test. It is not a real Microsoft exam and does not contain exam dumps or leaked exam content.

AI-901 practice test questions

Question 1: Which AI workload is best suited for grouping customers into segments based on similar purchasing behavior when no predefined labels are available?

• Regression
• Clustering
• Classification
• Object detection

The correct answer is Clustering.

Clustering is an unsupervised machine learning technique used to group similar records when the desired groups are not already labeled. It is useful for customer segmentation because the model identifies patterns from the data instead of learning from predefined categories.

Related Microsoft Learn topic

Fundamental machine learning concepts

Question 2: A model predicts whether an email is spam or not spam. Which machine learning task is being used?

• Regression
• Clustering
• Binary classification
• Anomaly detection

The correct answer is Binary classification.

Binary classification predicts one of two possible classes. In this case, each email is assigned to either spam or not spam, which makes the task a two-class classification problem.

Related Microsoft Learn topic

Fundamental machine learning concepts

Question 3: True or False: Regression models are commonly used to predict numeric values such as product demand, prices, or temperature.

• True
• False

The correct answer is true.

Regression models estimate continuous numeric values. They are appropriate when the expected output is a number, such as a sales forecast, delivery time estimate, or predicted cost.

Related Microsoft Learn topic

Fundamental machine learning concepts

Question 4: Which statements describe responsible AI principles. Choose 2 answers.

• AI systems should be designed to be fair and inclusive
• AI systems should always make decisions without human review
• AI systems should hide how decisions are made
• AI systems should protect privacy and security
• AI systems should ignore possible harm if accuracy is high

The correct answers are AI systems should be designed to be fair and inclusive and AI systems should protect privacy and security..

Responsible AI includes principles such as fairness, inclusiveness, privacy, security, transparency, reliability, safety, and accountability. These principles help teams build systems that are useful, explainable, and aligned with human values.

Related Microsoft Learn topic

Microsoft responsible AI principles

Question 5: Which computer vision capability identifies the location of multiple objects in an image and returns bounding boxes?

• Image classification
• Object detection
• Text summarization
• Sentiment analysis

The correct answer is Object detection.

Object detection identifies objects and their positions within an image, usually by returning bounding boxes. Image classification typically assigns one or more labels to the whole image without locating each object.

Related Microsoft Learn topic

Computer vision workloads

Question 6: A retail company wants to extract printed text from product labels in photos. Which AI capability should it use?

• Optical character recognition
• Speech synthesis
• Named entity recognition
• Forecasting

The correct answer is Optical character recognition.

Optical character recognition, or OCR, extracts text from images and scanned documents. It is the correct capability when text appears inside a visual file rather than as normal digital text.

Related Microsoft Learn topic

Read text in images

Question 7: True or False: Sentiment analysis can be used to estimate whether text expresses positive, negative, neutral, or mixed opinions.

• True
• False

The correct answer is true.

Sentiment analysis evaluates text to determine the emotional tone or opinion expressed. It is often used for customer reviews, support tickets, survey responses, and social feedback.

Related Microsoft Learn topic

Azure Language overview

Question 8: Which natural language processing capability identifies people, locations, organizations, dates, and quantities in text?

• Speech recognition
• Image tagging
• Object detection
• Named entity recognition

The correct answer is Named entity recognition.

Named entity recognition finds and categorizes important entities in text, such as people, places, organizations, dates, and numeric values. It helps applications transform unstructured text into structured information.

Related Microsoft Learn topic

Azure Language overview

Question 9: Which AI capability converts spoken audio into written text?

• Speech to text
• Text to speech
• Text classification
• Image analysis

The correct answer is Speech to text.

Speech to text transcribes spoken language into written text. This capability is used for captions, call center analytics, voice commands, and meeting transcription scenarios.

Related Microsoft Learn topic

Azure AI Speech overview

Question 10: Which AI capability converts written text into natural-sounding audio?

• Speech to text
• Text to speech
• Entity recognition
• Document classification

The correct answer is Text to speech.

Text to speech generates spoken audio from text input. It is commonly used for accessibility, voice assistants, interactive voice response systems, and audio content generation.

Related Microsoft Learn topic

Azure AI Speech overview

Question 11: Which statements are common characteristics of generative AI. Choose 2 answers.

• It can create new text, images, code, or other content
• It can only group rows of data into clusters
• It commonly uses prompts as input instructions
• It always returns deterministic output with no variation
• It requires all output to be manually prewritten

The correct answers are It can create new text, images, code, or other content and it commonly uses prompts as input instructions..

Generative AI creates new content based on user prompts or other input. It is often used for drafting text, summarizing information, generating images, producing code, and supporting conversational experiences.

Related Microsoft Learn topic

Generative AI fundamentals

Question 12: True or False: A prompt is an instruction or input given to a generative AI model to guide the model response.

• True
• False

The correct answer is true.

A prompt is the user-provided input that guides a generative AI model. Clear prompts can include instructions, context, constraints, examples, and the desired output format.

Related Microsoft Learn topic

Generative AI fundamentals

Question 13: What is grounding in the context of generative AI applications?

• Training a model only with images
• Removing all user context from a response
• Providing relevant source information so responses are based on trusted data
• Compressing a model to reduce latency

The correct answer is Providing relevant source information so responses are based on trusted data..

Grounding connects model responses to relevant data sources, such as documents, databases, or search results. It helps reduce unsupported answers by giving the model context that should be reflected in the response.

Related Microsoft Learn topic

Retrieval augmented generation concepts

Question 14: Which term describes incorrect or unsupported information generated by an AI model?

• Tokenization
• Classification
• Grounding
• Hallucination

The correct answer is Hallucination.

A hallucination occurs when a generative AI model produces content that sounds plausible but is factually incorrect, unsupported, or not grounded in the provided context. Grounding, evaluation, and human review help reduce this risk.

Related Microsoft Learn topic

Generative AI fundamentals

Question 15: A company wants an AI system to answer questions using its own policy documents. Which design pattern is commonly used?

• Image classification
• Retrieval augmented generation
• Speech synthesis
• Clustering

The correct answer is Retrieval augmented generation.

Retrieval augmented generation retrieves relevant content from a trusted knowledge source and provides it to a generative model as context. This pattern helps build question-answering systems over private or enterprise data.

Related Microsoft Learn topic

Retrieval augmented generation concepts

Question 16: Which statements describe AI agents. Choose 2 answers.

• Agents can use tools or actions to complete tasks
• Agents are limited to static image classification only
• Agents cannot use language models
• Agents must always be trained from scratch
• Agents can reason over instructions and context to support multi-step workflows

The correct answers are Agents can use tools or actions to complete tasks and agents can reason over instructions and context to support multi-step workflows..

AI agents commonly combine model reasoning, instructions, memory or context, and tools. They can perform multi-step tasks such as searching data, calling APIs, using knowledge sources, and producing task-specific outputs.

Related Microsoft Learn topic

Microsoft Foundry Agent Service

Question 17: True or False: Human review is unnecessary for all AI systems once a model reaches high accuracy in testing.

• True
• False

The correct answer is false.

Human review can still be required for sensitive, high-impact, or uncertain scenarios. Accuracy is only one quality measure; teams must also consider safety, fairness, privacy, reliability, and accountability.

Related Microsoft Learn topic

Microsoft responsible AI principles

Question 18: Which AI workload is most appropriate for detecting unusual credit card transactions?

• Anomaly detection
• Text translation
• Image classification
• Text to speech

The correct answer is Anomaly detection.

Anomaly detection identifies patterns that differ significantly from normal behavior. It is commonly used for fraud detection, unusual system telemetry, suspicious activity, and quality control.

Related Microsoft Learn topic

Anomaly detection concepts

Question 19: Which capability is most appropriate for translating customer support messages from one language to another?

• Object detection
• Regression
• Machine translation
• Speaker recognition

The correct answer is Machine translation.

Machine translation converts text from one language into another. It is useful when applications need multilingual support, localization, or cross-language communication.

Related Microsoft Learn topic

Azure AI Translator overview

Question 20: Which statements describe classification. Choose 2 answers.

• It predicts continuous numeric values only
• It assigns data to predefined categories
• It only works with unlabeled data
• It can be binary or multiclass
• It creates spoken audio from text

The correct answers are It assigns data to predefined categories and it can be binary or multiclass..

Classification assigns records to known categories. Binary classification predicts between two categories, while multiclass classification predicts one category from three or more possible classes.

Related Microsoft Learn topic

Fundamental machine learning concepts

Question 21: Which responsible AI principle is most closely related to explaining how an AI system makes decisions?

• Elasticity
• Transparency
• Object detection
• Token limit

The correct answer is Transparency.

Transparency is about making AI systems understandable to users, developers, and stakeholders. It supports explainability, clear communication, and appropriate trust in AI-powered decisions.

Related Microsoft Learn topic

Microsoft responsible AI principles

Question 22: Which platform is used in Azure to manage projects, deploy models, build agents, and monitor AI assets?

• Microsoft Foundry
• Azure Storage Explorer
• Azure DNS
• Microsoft Purview only

The correct answer is Microsoft Foundry.

Microsoft Foundry is the Azure platform for building, managing, deploying, and governing AI apps and agents. It provides access to models, projects, tools, agents, evaluation, and operational features for AI workloads.

Related Microsoft Learn topic

What is Microsoft Foundry?

Question 23: True or False: Microsoft Foundry can be used to build both generative AI applications and AI agents.

• True
• False

The correct answer is true.

Microsoft Foundry supports AI application and agent development. Teams can use it to select models, create deployments, configure tools, build agents, evaluate outputs, and manage AI assets.

Related Microsoft Learn topic

What is Microsoft Foundry?

Question 24: A developer needs to choose from available AI models and deploy one for use by an application. Which Microsoft Foundry capability is most relevant?

• Monthly archive pages
• Model catalog and model deployments
• Azure Blob lifecycle rules
• Azure DNS record sets

The correct answer is Model catalog and model deployments.

The model catalog helps users discover available models, while deployments make selected models available for application use. This is the normal flow for selecting a model and exposing it through an endpoint.

Related Microsoft Learn topic

Explore models in Microsoft Foundry

Question 25: Which object in Microsoft Foundry is commonly used to organize AI development work, resources, model deployments, and related assets?

• DNS zone
• Storage queue
• Foundry project
• Virtual network gateway

The correct answer is Foundry project.

A Foundry project provides an organized workspace for AI development. It helps teams manage related assets such as models, deployments, tools, evaluations, and agent work within a shared context.

Related Microsoft Learn topic

What is Microsoft Foundry?

Question 26: Which statements describe model deployment in Microsoft Foundry. Choose 2 answers.

• It makes a selected model available for application calls
• It automatically deletes all project assets
• It replaces the need for prompts in every scenario
• It can be associated with billing and consumption
• It is only used for DNS routing

The correct answers are It makes a selected model available for application calls and it can be associated with billing and consumption..

Deploying a model makes it usable by applications through an endpoint or managed experience. Consumption is typically associated with model usage, so teams should consider cost, quota, latency, and region availability.

Related Microsoft Learn topic

What is Microsoft Foundry?

Question 27: Which service provides a managed platform for building, deploying, and scaling AI agents in Microsoft Foundry?

• Azure Files
• Foundry Agent Service
• Azure Bastion
• Azure Monitor Workbooks only

The correct answer is Foundry Agent Service.

Foundry Agent Service is a managed platform for creating, deploying, and scaling AI agents. It supports agents that can use models, instructions, tools, and knowledge to complete tasks.

Related Microsoft Learn topic

Microsoft Foundry Agent Service

Question 28: True or False: Foundry Agent Service can use tools to extend what an agent can do beyond generating text.

• True
• False

The correct answer is true.

Agent tools allow agents to perform actions such as searching data, calling services, working with files, or connecting to knowledge. Tools extend the agent beyond simple text generation.

Related Microsoft Learn topic

Agent tools overview

Question 29: A team wants an agent to answer questions using enterprise documents while respecting access permissions. Which concept is most relevant?

• Object detection only
• Manual image tagging
• Text to speech only
• Permission-aware knowledge grounding

The correct answer is Permission-aware knowledge grounding.

Enterprise agents often need grounded knowledge from internal data sources while respecting user permissions. This helps the agent provide useful answers without exposing information that the user should not access.

Related Microsoft Learn topic

Foundry IQ

Question 30: Which Microsoft Foundry feature helps evaluate the quality and safety of generative AI outputs before production release?

• Evaluations
• DNS forwarding
• Static website hosting
• Blob soft delete

The correct answer is Evaluations.

Evaluations help teams measure AI application behavior, quality, safety, groundedness, and other important dimensions. Evaluation is a key step before deploying AI apps or agents to production users.

Related Microsoft Learn topic

Evaluate generative AI applications

Question 31: Which statements are good practices when creating prompts for a generative AI application. Choose 2 answers.

• Provide no context so the model guesses the goal
• Specify the desired output format
• Mix unrelated tasks in the same prompt without instructions
• Avoid giving constraints for sensitive tasks
• Include relevant context and clear instructions

The correct answers are Specify the desired output format and include relevant context and clear instructions..

Good prompts are specific, clear, and contextual. Including the desired format, constraints, examples, and relevant background helps the model produce more useful and predictable responses.

Related Microsoft Learn topic

Get started with generative AI

Question 32: Which Azure AI service in Foundry Tools provides NLP capabilities for understanding and analyzing text?

• Azure Load Balancer
• Azure Backup
• Azure Language
• Azure DNS

The correct answer is Azure Language.

Azure Language provides natural language processing capabilities such as entity recognition, sentiment analysis, key phrase extraction, language detection, and text analytics features.

Related Microsoft Learn topic

Azure Language overview

Question 33: A support team wants to identify key topics from thousands of customer comments. Which Azure Language capability is most appropriate?

• Key phrase extraction
• Image classification
• Speaker verification
• Object localization

The correct answer is Key phrase extraction.

Key phrase extraction identifies important words and phrases from text. It helps summarize large collections of feedback and discover common themes without manually reading every comment.

Related Microsoft Learn topic

Azure Language overview

Question 34: True or False: Azure Language can support applications built with REST APIs, client libraries, and the Microsoft Foundry web experience.

• True
• False

The correct answer is true.

Azure Language is available for application development through supported APIs and client libraries, and it can also be used from the Microsoft Foundry experience. This gives teams multiple ways to integrate NLP features.

Related Microsoft Learn topic

Azure Language overview

Question 35: Which Foundry Tool uses generative AI to process documents, images, audio, and video into structured output?

• Azure DNS
• Azure Queue Storage
• Azure Virtual WAN
• Azure Content Understanding

The correct answer is Azure Content Understanding.

Azure Content Understanding in Foundry Tools processes multiple content types, including documents, images, video, and audio. It can transform unstructured content into structured output defined by the user.

Related Microsoft Learn topic

Azure Content Understanding overview

Question 36: Which scenarios are suitable for Azure Content Understanding. Choose 2 answers.

• Extracting fields from documents into structured data
• Allocating static IP addresses to virtual machines
• Analyzing audio, images, or video for structured information
• Creating DNS zones for public domains
• Encrypting storage accounts only

The correct answers are Extracting fields from documents into structured data and analyzing audio, images, or video for structured information..

Content Understanding is designed to interpret unstructured or semi-structured content and convert it into useful structured information. It is appropriate for document extraction and multimodal content analysis scenarios.

Related Microsoft Learn topic

Azure Content Understanding overview

Question 37: A company needs to classify incoming documents and extract fields from them. Which Content Understanding concept is most relevant?

• Traffic routing
• Analyzer operation
• Virtual machine scale set
• Network security group

The correct answer is Analyzer operation.

Content Understanding analyzers can classify, split, and extract information from content. They are used to turn documents and other content into structured results that applications can consume.

Related Microsoft Learn topic

Content Understanding classification and segmentation

Question 38: Which Azure AI capability should be used to transcribe customer service calls into text for analysis?

• Speech to text
• Object detection
• Text to speech
• Document layout extraction

The correct answer is Speech to text.

Speech to text converts spoken audio into written text. Once transcribed, the text can be analyzed for sentiment, key phrases, entities, or other language insights.

Related Microsoft Learn topic

Azure AI Speech overview

Question 39: Which Azure AI capability should be used to generate voice output for an accessibility feature?

• Named entity recognition
• OCR
• Text to speech
• Clustering

The correct answer is Text to speech.

Text to speech converts text into spoken audio. It is appropriate for accessibility features, voice-enabled applications, guided experiences, and audio responses.

Related Microsoft Learn topic

Azure AI Speech overview

Question 40: True or False: A model deployment should be evaluated for cost, latency, quality, and safety before being used in a production AI application.

• True
• False

The correct answer is true.

Production AI systems should be tested for multiple operational and quality factors. Cost, latency, accuracy, groundedness, safety, and reliability all influence whether a model deployment is appropriate for the workload.

Related Microsoft Learn topic

Evaluate generative AI applications

Question 41: Which Azure AI capability extracts printed and handwritten text from images or scanned files?

• Speaker diarization
• Optical character recognition
• Regression
• Sentiment analysis

The correct answer is Optical character recognition.

Optical character recognition reads text from images and scanned content. It is useful for receipts, forms, labels, screenshots, and document digitization scenarios.

Related Microsoft Learn topic

Read text in images

Question 42: Which statements describe content safety controls for generative AI solutions. Choose 2 answers.

• They guarantee every response is always correct
• They can help detect or block harmful categories of content
• They replace the need for application testing
• They help reduce risk in user prompts and model outputs
• They are used only for image resizing

The correct answers are They can help detect or block harmful categories of content and they help reduce risk in user prompts and model outputs..

Content safety controls help identify and mitigate harmful user input and model output. They are part of a broader responsible AI approach that also includes grounding, evaluation, monitoring, and human oversight.

Related Microsoft Learn topic

Azure AI Content Safety overview

Question 43: A developer wants a chat application to answer only from approved support articles and say when it does not know. Which implementation approach is best?

• Allow the model to answer from general knowledge only
• Remove all source documents from the prompt
• Use image classification for every question
• Use grounding with approved knowledge sources and clear response instructions

The correct answer is Use grounding with approved knowledge sources and clear response instructions..

Grounding the model with approved support articles helps constrain responses to trusted content. Clear instructions can tell the model how to behave when the answer is not available in the provided sources.

Related Microsoft Learn topic

Retrieval augmented generation concepts

Question 44: Which tool capability allows an agent to retrieve current or external information instead of relying only on the model prompt?

• Search or knowledge tools
• DNS aliases
• Subnet delegation
• Blob immutable storage

The correct answer is Search or knowledge tools.

Search and knowledge tools allow agents to retrieve information from configured sources. This improves usefulness in scenarios where the answer depends on documents, enterprise knowledge, or current data.

Related Microsoft Learn topic

Agent tools overview

Question 45: True or False: AI agent instructions should define the agent role, boundaries, and expected behavior.

• True
• False

The correct answer is true.

Instructions help define what the agent should do, how it should behave, and what limits it should respect. Good instructions improve consistency, safety, and alignment with the intended task.

Related Microsoft Learn topic

Microsoft Foundry Agent Service

Question 46: Which statements are useful when evaluating AI-generated answers. Choose 2 answers.

• Check whether the answer is grounded in the provided sources
• Ignore incorrect answers if they sound fluent
• Measure only response length
• Avoid testing with realistic user prompts
• Review quality, safety, and relevance

The correct answers are Check whether the answer is grounded in the provided sources and review quality, safety, and relevance..

AI-generated answers should be evaluated for quality, relevance, safety, and groundedness. Fluent wording does not guarantee correctness, so realistic test prompts and source validation are important.

Related Microsoft Learn topic

Evaluate generative AI applications

Question 47: Which capability is most appropriate for identifying the language used in a customer message before routing it to the correct support queue?

• Text to speech
• Object detection
• Language detection
• Face detection

The correct answer is Language detection.

Language detection identifies the language of text. It is useful before translation, localization, routing, or applying language-specific processing in a support workflow.

Related Microsoft Learn topic

Azure Language overview

Question 48: A business wants to summarize long policy documents for employees. Which AI workload is most relevant?

• Object detection
• Text summarization
• Speaker identification
• Image resizing

The correct answer is Text summarization.

Text summarization condenses long text into shorter output while preserving important points. It is useful for policies, reports, support cases, meeting transcripts, and research documents.

Related Microsoft Learn topic

Azure Language overview

Question 49: Which statements describe safe deployment of AI applications. Choose 2 answers.

• Skip monitoring after release
• Use production users as the only test group
• Monitor behavior and collect feedback after deployment
• Remove all content filters to improve speed
• Evaluate outputs before release and continue improving the system

The correct answers are Monitor behavior and collect feedback after deployment and evaluate outputs before release and continue improving the system..

Safe AI deployment includes testing before release and monitoring after release. Teams should collect feedback, evaluate real behavior, review safety signals, and improve prompts, grounding, or application logic as needed.

Related Microsoft Learn topic

Evaluate generative AI applications

Question 50: Which Azure AI service helps build applications that can detect and filter harmful user-generated content?

• Azure AI Content Safety
• Azure DNS
• Azure File Sync
• Azure Load Balancer

The correct answer is Azure AI Content Safety.

Azure AI Content Safety helps detect harmful content in user input and AI-generated output. It supports responsible AI implementation by reducing the risk of harmful or inappropriate content.

Related Microsoft Learn topic

Azure AI Content Safety overview