Questions & answers (free, no sign-in needed)
- View
10 beginner questions on the AI-901 skills measured: responsible AI principles, model configuration (temperature), AI workloads (speech, text analysis), and implementing solutions with Microsoft Foundry: system prompts, agents, Azure Language, image generation and Azure Content Understanding.
Ready to test yourself? Take the graded Beginner test (free; sign in with Google, one click) to get your score.
A company tests a loan-approval model and finds that it approves equally qualified applicants from one group far less often than others. Which responsible AI principle is most directly at risk?
- A. Reliability and safety
- B. Fairness ✔ correct
- C. Transparency
- D. Privacy and security
Answer: B. Fairness means AI systems should treat everyone fairly and avoid affecting similar groups differently, for example in loan decisions for people with similar qualifications. Reliability and safety is about consistent, safe operation.
Source: learn.microsoft.com/en-us/azure/machine-learning/concept-responsible-ai · Domain: Describe principles of responsible AI
Under the accountability principle, the people who design and deploy an AI system are responsible for how that system operates.
- A. True ✔ correct
- B. False
Answer: A. True. Microsoft's accountability principle states that people who design and deploy AI systems must be accountable for how those systems operate, supported by governance, oversight and industry standards.
Source: learn.microsoft.com/en-us/azure/machine-learning/concept-responsible-ai · Domain: Describe principles of responsible AI
For a chat model that supports the temperature parameter, what is the usual effect of lowering the temperature (for example, from 0.7 to 0.2)?
- A. Responses become more focused and deterministic ✔ correct
- B. Responses become more random and creative
- C. The model can read longer prompts
- D. The model answers in more languages
Answer: A. A lower temperature makes output more focused and deterministic, while a higher value makes it more random. Temperature does not change the context length or the languages a model supports.
Source: learn.microsoft.com/en-us/azure/foundry/openai/concepts/prompt-engineering · Domain: Identify AI model components and configurations
A company wants written transcripts of its recorded customer support calls. Which AI workload does this require?
- A. Speech synthesis (text to speech)
- B. Image classification
- C. Speech recognition (speech to text) ✔ correct
- D. Sentiment analysis
Answer: C. Converting spoken audio into text is speech recognition, provided by the speech to text capability of Azure Speech. Speech synthesis does the opposite and turns text into spoken audio.
Source: learn.microsoft.com/en-us/azure/ai-services/speech-service/overview · Domain: Identify AI workloads
Which two are text analysis techniques? (Select TWO)
- A. Key phrase extraction ✔ correct
- B. Named entity recognition ✔ correct
- C. Speech synthesis
- D. Image generation
- E. Optical character recognition
Answer: A, B. Key phrase extraction returns the main concepts in text, and named entity recognition finds entities such as people, places and organizations. Both are features of Azure Language; the others work with speech or images.
Source: learn.microsoft.com/en-us/azure/ai-services/language-service/overview · Domain: Identify AI workloads
A developer is building a chat app on a model deployed in Microsoft Foundry. Every answer must be polite and limited to questions about the company's products. Where should these instructions go?
- A. In the deployment name of the model
- B. In each user prompt typed by end users
- C. In the maximum token limit setting
- D. In the system message (system prompt) ✔ correct
Answer: D. The system message gives the model instructions and context that apply to the whole conversation, such as its role, scope, tone and output format. End users should not have to repeat these rules in their prompts.
Source: learn.microsoft.com/en-us/azure/foundry/openai/concepts/advanced-prompt-engineer · Domain: Implement generative AI apps and agents by using Foundry
In Foundry Agent Service, what do you configure to create a prompt agent?
- A. A model, instructions and the tools the agent can use ✔ correct
- B. A virtual machine, an operating system and a web server
- C. A labeled training dataset and a training schedule
- D. A storage account, a container and a lifecycle rule
Answer: A. A prompt agent is defined by a model, instructions and tools (such as file search or code interpreter), and Foundry runs it without your own code or infrastructure. No model training or VMs are needed.
Source: learn.microsoft.com/en-us/azure/foundry/agents/overview · Domain: Implement generative AI apps and agents by using Foundry
A support team wants its app to label each customer review as positive, neutral or negative. Which Foundry Tools capability should it call?
- A. Text to speech in Azure Speech
- B. Sentiment analysis in Azure Language ✔ correct
- C. Image generation with an image model
- D. Text translation with Azure Translator
Answer: B. Sentiment analysis in Azure Language assigns labels such as positive, neutral and negative to sentences and whole documents, with confidence scores. The other capabilities produce audio, images or translated text.
Source: learn.microsoft.com/en-us/azure/ai-services/language-service/sentiment-opinion-m · Domain: Implement AI solutions for text and speech by using Foundry
A marketing team wants to create new product images from short text descriptions. Which kind of model should it deploy in Microsoft Foundry?
- A. A text embedding model
- B. A speech to text model
- C. An image generation model ✔ correct
- D. A sentiment analysis model
Answer: C. Image generation models create new images from text prompts (and can edit existing images). Embedding models turn text into vectors, and speech to text models transcribe audio; neither produces images.
Source: learn.microsoft.com/en-us/azure/foundry/openai/how-to/dall-e · Domain: Implement AI solutions with computer vision and image-generation capabilities by using Foundry
A company wants one service to extract structured fields from invoices and also from recorded audio and video files. Which Foundry Tools service fits best?
- A. Azure Translator
- B. Text to speech in Azure Speech
- C. A text embedding model
- D. Azure Content Understanding ✔ correct
Answer: D. Azure Content Understanding in Foundry Tools uses generative AI to process documents, images, audio and video into a user-defined output format, such as invoice fields. Translator only translates text.
Source: learn.microsoft.com/en-us/azure/ai-services/content-understanding/overview · Domain: Implement AI solutions for information extraction by using Foundry
Original practice questions written by xOps School (DevOpsSchool Private Limited) from the public Microsoft exam objectives (Skills measured as of 2026-04-15 (AI-901 replaced AI-900, retired 2026-06-30); 7 skill areas in 2 groups: AI concepts 40–45%, Microsoft Foundry 55–60%), checked 2026-10-25. They are not actual exam questions. xOps School is not affiliated with or endorsed by Microsoft; exam names are trademarks of their owners. Official exam page: learn.microsoft.com/en-us/credentials/certifications/resources/study-g.