10 free Microsoft AI-901 practice questions with the correct answer and a full explanation for each, taken from the CertStash pack of 128 questions. Work through them, then open each answer to check your reasoning.
Get all 128 questions (US$39) · Download these 10 as a PDF
Question 1
You are using the Azure Speech SDK to develop a Python application that supports rea-ime spoken conversations.
Which Azure Speech class should you use to configure the connection to the Azure Speech service?
Show answer and explanation
Correct answer: D. SpeechConfig
SpeechConfig is the correct class to configure the connection parameters to the Azure Speech service, including subscription key, region, and authentication details. It establishes the foundational connection settings needed before creating speech recognizers or synthesizers.
Why the other options are wrong
- A. AudioOutputConfig specifies output audio settings but does not configure the connection to the Azure Speech service itself.
- B. SpeechSynthesizer is used for text-to-speech synthesis after the connection is already configured.
- C. AudioConfig handles input/output audio device configuration, not service connection setup.
Question 2
You need to create an AI agent in Microsoft Foundry that follows a specific role and behavior when responding to users.
What should you configure?
Show answer and explanation
Correct answer: A. system instructions
System instructions define the specific role, persona, and behavioral guidelines that an AI agent follows when responding to users. These instructions shape how the agent interprets requests and generates responses aligned with desired conduct.
Why the other options are wrong
- B. Temperature controls response randomness and creativity, not the agent's role or behavior definition.
- C. Tokens per minute (TPM) is a rate-limiting parameter for API usage, not behavioral configuration.
- D. Max completion tokens limits response length but does not define role or behavior.
Question 3
You are developing a web app that processes invoices to calculate expenses.
You need to extract structured fields, including nested values, from the invoices by using a defined schema.
What should you use?
Show answer and explanation
Correct answer: C. an analyzer in Azure Content Understanding in Foundry Tools
Foundry Tools An analyzer in Azure Content Understanding in Foundry Tools is designed to extract structured fields, including nested values, from documents using a defined schema. This tool supports sophisticated document processing with schema-driven extraction.
Why the other options are wrong
- A. OCR-only pipelines extract text but cannot structure data according to a schema or handle nested values.
- B. Azure Speech transcription workflows are designed for audio-to-text conversion, not document field extraction.
- D. Azure AI Search is a search and retrieval service, not a document structure extraction tool.
Question 4
What are two purposes of instructions when prompting a generative AI model? Each correct answer presents part of the solution.
NOTE: Each correct selection is worth one point.
Show answer and explanation
Correct answer: A, E
A. defines the agent’s role and behavior E. defines constraints on the model’s responses Instructions serve two critical purposes: they define the agent's role and behavior (what persona the model should adopt and how it should act), and they define constraints on the model's responses (boundaries, output format, and restrictions on what the model should or should not do).
Why the other options are wrong
- B. Model selection is determined by deployment configuration or API parameters, not by instructions within a prompt.
- C. Azure region is set during service provisioning or deployment configuration, independent of prompt instructions.
- D. TPM allocation is configured at the service level for rate limiting, not through prompt instructions.
Question 5
You need to build an AI solution that produces new product images based on written descriptions provided by users.
Which AI workload should you use?
Show answer and explanation
Correct answer: B. image generation
Image generation is the AI workload that produces new, original images based on textual descriptions. This workload uses generative models to create images that match the provided specifications.
Why the other options are wrong
- A. Image analysis examines existing images to extract information, rather than generating new ones.
- C. Object detection identifies and locates objects within existing images, not generating new images.
- D. Optical character recognition (OCR) extracts text from images, unrelated to image generation.
Question 6
You have a Microsoft Foundry project that has a generative AI model deployment.
You need to ensure that responses generated by the model minimize costs and remain within a defined length.
Which parameter should you configure?
Show answer and explanation
Correct answer: C. Max Completion Tokens
Max Completion Tokens controls the maximum length of generated responses by limiting the number of tokens the model can produce. This directly minimizes costs by reducing processing load and ensures responses stay within defined length boundaries.
Why the other options are wrong
- A. Top P is a sampling parameter that affects response diversity, not response length or cost.
- B. Temperature controls randomness in responses, not length or token usage.
- D. Model version settings determine which model variant to use, not response length constraints.
Question 7
Your company processes customer support emails.
You need to implement an AI solution that automatically identifies mentions of people, organizations, and locations in the emails.
Which text analysis technique should you use?
Show answer and explanation
Correct answer: A. Named Entity Recognition (NER)
Named Entity Recognition (NER) is specifically designed to identify and classify named entities such as people, organizations, locations, and other proper nouns within text. This is the standard technique for extracting entity mentions from unstructured documents.
Why the other options are wrong
- B. Key phrase extraction identifies important topics and phrases but does not specifically categorize people, organizations, and locations.
- C. Sentiment analysis determines emotional tone or opinion in text, unrelated to entity identification.
- D. Summarization condenses text content rather than identifying specific entity mentions.
Question 8
You need to build an AI solution that generates marketing email drafts based on a short description of a product and its target audience.
Which AI workload should you use?
Show answer and explanation
Correct answer: C. generative AI
Generative AI is the workload that creates new content, such as marketing email drafts, based on input descriptions and parameters. This workload uses language models to generate contextually relevant and original text.
Why the other options are wrong
- A. Speech recognition converts spoken audio to text, unrelated to marketing email generation.
- B. Text classification categorizes existing text into predefined categories, not generating new content.
- D. Computer vision processes and analyzes images, not relevant to generating tex-ased marketing emails.
Question 9
You are developing an AI-powered customer support application.
Which task is an example of the Microsoft responsible AI principle of inclusiveness?
Show answer and explanation
Correct answer: C. Design the interface to support multiple languages and screen readers.
Inclusiveness in Microsoft's responsible AI framework focuses on ensuring AI systems are accessible and usable by everyone, regardless of ability or background. Designing interfaces to support multiple languages and screen readers directly addresses accessibility for users with different needs and linguistic backgrounds, making the technology inclusive to broader populations.
Why the other options are wrong
- A. Data encryption and role-based access control relate to security and privacy principles, not inclusiveness.
- B. Evaluating model outputs across demographic groups addresses fairness and bias reduction, which is the fairness principle, not inclusiveness.
- D. Providing explanations about predictions relates to transparency and explainability principles, not inclusiveness.
Question 10
You have a Microsoft Foundry project that contains a generative AI model deployment.
You test the model by using the Foundry playground.
You need to develop an application that sends requests to the deployed model.
Which information must the application include to call the model?
Show answer and explanation
Correct answer: D. The model endpoint and authentication credentials
To call a deployed model programmatically, an application must have the model endpoint URL where the service is hosted and authentication credentials (such as API keys or tokens) to authorize requests. These are the fundamental technical requirements for establishing communication with any deployed service. The playground session history, training dataset, and project display name are not required for making API calls to a deployed model.
Why the other options are wrong
- A. Playground session history is useful for testing but not required for application integration with the deployed model.
- B. The training dataset is not needed to call an already-deployed model; the model itself is what matters.
- C. The project display name is a UI identifier and has no role in API authentication or endpoint communication.
That was 10 of 128.
The full Microsoft AI-901 pack has all 128 questions, each with the answer, the explanation and why the other options are wrong, plus a questions-only copy for timed runs. US$39, paid once, with free monthly updates and a pass-or-your-money-back guarantee.