Browse Certification Practice Tests by Exam Family

AI-900: Generative AI

Try 10 focused AI-900 questions on Generative AI, with explanations, then continue with IT Mastery.

On this page

Open the matching IT Mastery practice page for timed mocks, topic drills, progress tracking, explanations, and full practice.

Try AI-900 on Web View full AI-900 practice page

Topic snapshot

FieldDetail
Exam routeAI-900
Topic areaDescribe Features of Generative AI Workloads on Azure
Blueprint weight24%
Page purposeFocused sample questions before returning to mixed practice

How to use this topic drill

Use this page to isolate Describe Features of Generative AI Workloads on Azure for AI-900. Work through the 10 questions first, then review the explanations and return to mixed practice in IT Mastery.

PassWhat to doWhat to record
First attemptAnswer without checking the explanation first.The fact, rule, calculation, or judgment point that controlled your answer.
ReviewRead the explanation even when you were correct.Why the best answer is stronger than the closest distractor.
RepairRepeat only missed or uncertain items after a short break.The pattern behind misses, not the answer letter.
TransferReturn to mixed practice once the topic feels stable.Whether the same skill holds up when the topic is no longer obvious.

Blueprint context: 24% of the practice outline. A focused topic score can overstate readiness if you recognize the pattern too quickly, so use it as repair work before timed mixed sets.

Sample questions

These questions are original IT Mastery practice items aligned to this topic area. They are designed for self-assessment and are not official exam questions.

Question 1

Topic: Describe Features of Generative AI Workloads on Azure

A company wants to build a generative AI assistant that summarizes support articles and answers employee questions. Before choosing a model, the team wants to discover available models in Azure and compare them because not every model will fit the use case. Which Azure capability is the best starting point?

Options:

  • A. Azure AI Vision

  • B. Azure OpenAI Service

  • C. Azure AI Foundry model catalog

  • D. Azure AI Language

Best answer: C

Explanation: Azure AI Foundry model catalog is used to discover and compare model options for generative AI scenarios. It supports choosing a model based on the workload, rather than assuming every available model is a fit for summarization and question answering.

The core concept is model discovery and comparison. When a team has a generative AI use case but has not selected a model yet, a model catalog helps them review available choices and compare them against their needs. In this scenario, the team wants a chatbot-style assistant that can summarize content and answer questions, so Azure AI Foundry model catalog is the best starting point.

A model catalog does not imply that every listed model is suitable for every use case. Teams still need to evaluate which model best matches requirements such as:

  • chat and question-answer behavior
  • summarization capability
  • overall fit for the business scenario

A generative AI service can host or provide models, but the stem specifically asks for the place to discover and compare options first.

  • The option choosing Azure OpenAI Service is plausible for building generative AI solutions, but it does not directly match the need to first discover and compare model choices.
  • The option choosing Azure AI Language fits text analytics workloads such as sentiment analysis or entity extraction, not browsing generative model options.
  • The option choosing Azure AI Vision fits image analysis workloads, not selecting a text generation model for an assistant.

Question 2

Topic: Describe Features of Generative AI Workloads on Azure

A company plans to use Azure OpenAI Service for several tasks. For which task is human-in-the-loop review most important before the generated output is given to end users?

Options:

  • A. Drafting patient discharge instructions

  • B. Generating alternative titles for blog articles

  • C. Suggesting taglines for a seasonal ad campaign

  • D. Creating first-pass summaries of internal meetings

Best answer: A

Explanation: Human-in-the-loop review is most important when generated content could directly affect a person’s health, rights, or safety. Patient discharge instructions are high impact, so a human reviewer should check them before they are used.

Human-in-the-loop review means a person checks or approves AI-generated output before it is used. In generative AI, this safeguard becomes especially important for high-stakes scenarios where errors could harm people or materially affect outcomes. Patient discharge instructions are a clear example because incorrect, incomplete, or unsafe wording could create medical risk. A qualified clinician should review that content before it reaches a patient.

Creative or low-risk drafting tasks can still benefit from editing, but they usually do not require the same level of safeguard. The key idea is to apply stronger human oversight when the potential impact on people is higher.

  • Ad taglines may need brand review, but they usually are not safety-critical.
  • Meeting summaries can be checked for accuracy, but they are typically lower risk than medical guidance.
  • Blog titles are editorial content and usually do not create direct health or safety consequences.

Question 3

Topic: Describe Features of Generative AI Workloads on Azure

A retail company wants to build a generative AI assistant that answers employee questions by summarizing policy documents. The team wants to browse foundation models, test prompts, and compare responses before selecting one. They do not need to train a traditional predictive model. Which Azure offering is the BEST choice?

Options:

  • A. Azure AI Foundry

  • B. Azure Machine Learning

  • C. Azure AI Language

  • D. Azure OpenAI Service

Best answer: A

Explanation: Azure AI Foundry is intended for building generative AI solutions, including exploring foundation models, testing prompts, and evaluating outputs. That matches the assistant and model-comparison needs better than a traditional ML platform or a single-task AI service.

Azure AI Foundry is the best fit when the scenario is centered on generative AI solution building and model exploration. Here, the team wants to browse foundation models, try prompts, compare outputs, and build an assistant that summarizes company documents. That aligns with Azure AI Foundry and its model catalog experience. Azure Machine Learning is the broader platform for machine learning model development and lifecycle management, so it is not the best fundamentals-level choice when the need is primarily generative AI exploration. Azure AI Language is for prebuilt language analysis tasks, and Azure OpenAI Service provides access to OpenAI models but does not match the broader model exploration focus as directly. The key clue is generative assistant plus model comparison, not classic ML training.

  • Classic ML focus The option using Azure Machine Learning fits broader model training and lifecycle management more than beginner-level generative AI assistant exploration.
  • Prebuilt NLP The option using Azure AI Language is for tasks such as sentiment analysis or entity extraction, not broad generative model comparison.
  • Closer but narrower The option using Azure OpenAI Service supports generative AI, but the stem emphasizes wider model exploration and solution-building capabilities.

Question 4

Topic: Describe Features of Generative AI Workloads on Azure

Which statement correctly describes Azure AI Foundry?

Options:

  • A. It provides prebuilt image analysis and OCR capabilities.

  • B. It is only for training custom predictive models.

  • C. It converts speech to text and text to speech.

  • D. It combines model selection and AI solution development in one experience.

Best answer: D

Explanation: Azure AI Foundry is designed to give teams a unified place to work with generative AI. It brings together model discovery and selection with tools to build and evaluate AI solutions, instead of focusing on a single workload such as vision or speech.

Azure AI Foundry is a high-level Azure experience for generative AI development. For AI-900, the main idea is that it brings two things together: access to models through a model catalog and capabilities to develop solutions that use those models. This makes it different from a single-purpose AI service. It is meant to help teams explore model options and then build, test, and manage AI solutions in one place. By contrast, Azure AI Vision and Azure AI Speech focus on specific workloads, and Azure Machine Learning is broader machine learning tooling rather than this unified generative AI experience. The key takeaway is unified model choice plus solution development.

  • The prebuilt image-analysis idea matches Azure AI Vision, not Azure AI Foundry.
  • The speech conversion idea matches Azure AI Speech capabilities.
  • The training-only idea is too narrow and is closer to Azure Machine Learning than to Azure AI Foundry.

Question 5

Topic: Describe Features of Generative AI Workloads on Azure

Which statement about generative AI output is correct?

Options:

  • A. It only returns text that appeared exactly in its training data.

  • B. It guarantees factually correct answers when the prompt is clear.

  • C. It can answer only questions that have a single correct response.

  • D. It can sometimes generate plausible but incorrect content, called hallucination.

Best answer: D

Explanation: Generative AI can produce fluent responses that seem believable even when they are wrong. This risk is called hallucination, so outputs may need verification or grounding in trusted data.

A core fact about generative AI is that it generates content from learned patterns rather than guaranteeing factual retrieval. Because of that, a model can sometimes produce an answer that is coherent and confident but still incorrect, unsupported, or completely invented. This behavior is known as hallucination.

In AI-900, you should recognize hallucination as an important limitation of generative AI solutions. It is one reason organizations often add human review, grounding on trusted sources, and safety controls. A well-written prompt may improve the response, but it does not guarantee truth. The key takeaway is that natural-sounding output is not the same as verified accuracy.

  • Guaranteed accuracy fails because a clear prompt can improve relevance but does not eliminate fabricated or incorrect answers.
  • Exact training text fails because generative AI can create new wording instead of only repeating memorized text.
  • Single-answer only fails because generative AI also supports open-ended tasks such as drafting, summarizing, and brainstorming.

Question 6

Topic: Describe Features of Generative AI Workloads on Azure

A healthcare provider uses Azure OpenAI Service for several tasks. For which task is clear human oversight most necessary before the generated response is used with patients or staff?

Options:

  • A. Drafting medication dosage guidance for patients

  • B. Creating titles for internal training articles

  • C. Rewriting appointment reminders in a friendlier tone

  • D. Suggesting taglines for a wellness campaign

Best answer: A

Explanation: Generative AI needs the strongest human oversight when its output could affect health, safety, legal outcomes, or finances. Medication dosage guidance is a high-impact use case, so human review is essential before the content is used.

A key generative AI concept is that oversight should increase as the risk of the output increases. Generative models can produce fluent text that sounds correct even when it is incomplete or wrong. For low-risk creative tasks, such as brainstorming titles or taglines, review is still helpful, but mistakes usually have limited impact. For high-impact tasks, such as patient dosage guidance, the output can directly affect safety and should be checked by a qualified person before anyone acts on it. In many cases, grounding the model with trusted source content can also help reduce unsupported answers, but human review remains especially important for sensitive decisions.

  • Creative marketing is lower risk; review helps quality, but errors usually do not directly endanger patients.
  • Tone changes for appointment reminders are mainly stylistic and typically have less impact than clinical guidance.
  • Training article titles are brainstorming output that can be corrected later without the same safety consequences.

Question 7

Topic: Describe Features of Generative AI Workloads on Azure

A company uses Azure OpenAI Service to answer employee questions from an internal benefits guide. The guide states that employees receive medical, dental, and vision coverage and 15 vacation days per year. It does not mention any sabbatical program. Which assistant response most clearly shows a hallucination?

Options:

  • A. The guide includes medical, dental, and vision coverage.

  • B. Employees receive a four-week paid sabbatical after five years.

  • C. Please specify whether you want medical, dental, or vision details.

  • D. Employees receive 15 vacation days each year.

Best answer: B

Explanation: A hallucination happens when generative AI produces content that sounds plausible but is unsupported or false. The sabbatical reply adds a policy that the guide does not contain, so it is the clearest example of hallucination.

Generative AI models create likely next words based on patterns, so they can sometimes produce confident-sounding answers that are incorrect or fabricated. That behavior is commonly called a hallucination. In this scenario, the assistant claims employees receive a paid sabbatical even though the benefits guide explicitly does not mention any sabbatical program. By contrast, repeating facts from the guide or asking a clarifying question does not introduce unsupported information. The key sign of hallucination is invented content, not simply reworded or shortened content.

  • The option restating 15 vacation days matches the provided guide, so it is supported.
  • The option listing medical, dental, and vision coverage also matches the source content.
  • The option asking for clarification avoids making up facts and is not a hallucination.

Question 8

Topic: Describe Features of Generative AI Workloads on Azure

A product team wants to build a generative AI assistant in Azure. Before choosing a model, they want to discover available model options and compare them at a high level. They also know that a listed model might still be a poor fit for their specific task. Which Azure capability best supports this step?

Options:

  • A. Azure AI Language

  • B. Azure OpenAI Service deployment

  • C. Azure AI Foundry model catalog

  • D. Azure Machine Learning automated ML

Best answer: C

Explanation: Azure AI Foundry model catalog helps teams discover and compare foundation models before selecting one to evaluate further. It supports exploration, but the team must still decide whether a model fits the task, quality, cost, and responsible AI needs of the specific use case.

The key concept is that a model catalog is a discovery and comparison tool, not a promise that every available model is appropriate for every business problem. In Azure, Azure AI Foundry model catalog lets a team review available generative AI model options and compare them at a high level before testing the most suitable candidates.

  • Use it to find candidate models.
  • Compare model choices for a scenario such as an assistant or summarization tool.
  • Then evaluate whether a selected model fits the task, constraints, and responsible AI requirements.

By contrast, a service that runs a chosen model or a specialized AI service for a specific workload does not primarily serve as a catalog for broad model discovery and comparison.

  • Specific deployment focuses on using a selected model, not on browsing and comparing a wider set of model options.
  • Specialized NLP service provides prebuilt language capabilities rather than a catalog of generative foundation models.
  • Predictive ML tooling is aimed at building or training machine learning models from data, not discovering foundation models for generative AI use cases.

Question 9

Topic: Describe Features of Generative AI Workloads on Azure

A company plans to use Azure OpenAI Service to draft new product descriptions from a short list of features. Which statement best describes a generative AI model?

Options:

  • A. It finds similar records by grouping unlabeled data.

  • B. It assigns each input to a predefined category.

  • C. It extracts existing entities and key phrases from text.

  • D. It creates new content based on learned patterns.

Best answer: D

Explanation: Generative AI models are designed to produce new outputs, such as writing text from a prompt. In this scenario, drafting product descriptions from feature lists is content creation, which is the defining characteristic of generative AI.

The key idea is that generative AI creates new content rather than only analyzing existing content. A model used to draft product descriptions from feature lists is generating original text based on patterns it learned during training. That makes it different from models that classify data into fixed labels, extract information already present in text, or cluster similar items.

In AI-900 terms, generative AI workloads include tasks like:

  • writing text
  • summarizing or rewriting content
  • generating images or code

A useful test is to ask whether the model is producing something new. If the output is newly created content, the workload is generative AI; if it only tags, extracts, or groups data, it is not.

  • Classification only fits tasks like spam detection or image labeling, not creating new descriptions.
  • Extraction only applies to pulling entities or phrases already present in source text.
  • Clustering only groups similar unlabeled items and does not generate content.

Question 10

Topic: Describe Features of Generative AI Workloads on Azure

A company wants to add a copilot-style assistant to its employee portal. The assistant must hold natural conversations and generate summaries of long policy documents. Which Azure service is the most appropriate choice?

Options:

  • A. Azure AI Speech

  • B. Azure Machine Learning

  • C. Azure OpenAI Service

  • D. Azure AI Language

Best answer: C

Explanation: This scenario describes a generative AI workload: chat plus generated summaries. Azure OpenAI Service is designed for using large language models in Azure for conversational assistants and text generation tasks such as summarization.

The key distinction here is between analyzing text and generating new text. A copilot-style assistant that chats with users and produces summaries is a generative AI solution, so the best fit is Azure OpenAI Service. It provides access to large language models that can create conversational responses, summarize content, and assist users in natural language.

Azure AI Language is primarily used for NLP analysis tasks such as sentiment analysis, entity recognition, and other text understanding features. Azure AI Speech is for spoken audio scenarios like speech-to-text and text-to-speech. Azure Machine Learning is a broader platform for building and managing custom machine learning solutions, but it is not the most direct choice for a basic generative assistant requirement.

The deciding clue is the need for generated chat responses, not just text analysis.

  • Text analysis confusion The Azure AI Language option is better for analyzing or extracting information from text than for a copilot-style generative chat experience.
  • Audio workload mismatch The Azure AI Speech option fits spoken language scenarios, not document-based chat and summarization.
  • Custom platform confusion The Azure Machine Learning option can support custom AI solutions, but it is less direct than a purpose-fit generative AI service for this requirement.

Continue with full practice

Use the AI-900 Practice Test page for the full IT Mastery route, mixed-topic practice, timed mock exams, explanations, and web/mobile app access.

Try AI-900 on Web View AI-900 Practice Test

Free review resource

Read the AI-900 Cheat Sheet on Tech Exam Lexicon, then return to IT Mastery for timed practice.

Revised on Thursday, May 14, 2026