About these questions
These five questions are original. We wrote them to test the same kinds of ideas as the AI-901 skills measured, with the two domains in rough proportion to their weights. They are not real exam questions, and the real exam will word things differently.
How to use these
Try each question without notes, and choose an answer before you open the explanation. Then read the explanation even when you were right, because the reason the closest wrong option fails is where the learning is. If you slip on a topic, revisit it on the domains explained page.
Question 1 (Identify AI concepts and capabilities)
A hotel group wants to scan thousands of guest reviews and list the main talking points in each one, such as “late checkout”, “friendly staff”, and “breakfast buffet”. The talking points do not need to be labeled with a type or category. Which text analysis technique fits?
- A. Entity recognition
- B. Language detection
- C. Key phrase extraction
- D. Sentiment analysis
Show the answer and explanation
Correct answer: C. Key phrase extraction returns the main concepts in a text as short phrases, with no type attached. Entity recognition is the closest wrong option. It finds items such as people, places, and organizations and labels each with a category, which the hotel group does not want. Language detection only says which language a text is written in. Sentiment analysis rates how positive or negative the text is, so it would not list topics.
Question 2 (Implement AI solutions by using Microsoft Foundry)
A university records a single 50-minute lecture as one audio file. A developer needs the full transcript back from one synchronous call, quicker than the length of the recording, with the different speakers told apart. Which Azure Speech capability should the developer use?
- A. Speech translation
- B. Real-time speech to text from a microphone
- C. Fast transcription
- D. Batch transcription
Show the answer and explanation
Correct answer: C. Fast transcription takes an audio file and returns the transcript synchronously, faster than real time, and it can separate speakers. Batch transcription also works on stored files, which makes it the closest wrong option, but it runs asynchronously and does not say when it will finish, so it does not meet the “one synchronous call” need. Real-time speech to text is designed for streaming input, such as a live microphone. Speech translation converts speech into another language, which was not requested.
Question 3 (Identify AI concepts and capabilities)
A software company is writing guidelines for its AI products. Which two of the following are Microsoft responsible AI principles? Choose two.
- A. Scalability
- B. Fairness
- C. Cost optimization
- D. Inclusiveness
- E. Latency
Show the answer and explanation
Correct answers: B and D. Fairness and inclusiveness are two of Microsoft’s six responsible AI principles. The others are reliability and safety, privacy and security, transparency, and accountability. Scalability and cost optimization are the closest wrong options because they sound like good engineering goals, but they come from cloud architecture guidance, not from the responsible AI principles. Latency is a performance measure and is not a principle.
Question 4 (Implement AI solutions by using Microsoft Foundry)
A ferry company is building an agent in Microsoft Foundry to help passengers plan trips. The agent must answer questions about today’s sea conditions and live service disruptions at the destination. Which tool is the most appropriate to add?
- A. A code interpreter
- B. File search over uploaded timetable documents
- C. A web search or web grounding tool
- D. A second model deployment
Show the answer and explanation
Correct answer: C. Sea conditions and live disruptions change throughout the day, so the agent needs a tool that can look up current information on the web and return it. File search is the closest wrong option. It only searches documents you uploaded, so it can answer questions about the timetable but cannot know today’s conditions unless someone keeps uploading fresh files. A code interpreter runs code and has no way to browse. Adding a second model gives more reasoning power but no access to live data.
Question 5 (Implement AI solutions by using Microsoft Foundry)
A legal team builds a Content Understanding analyzer for supplier contracts. One field must hold a two-sentence plain-language summary of the termination clause. The summary text does not appear in the contract word for word. Which field extraction method should the team choose?
- A. Extract
- B. Classify
- C. Optical character recognition
- D. Generate
Show the answer and explanation
Correct answer: D. The generate method produces a value the model infers from the content, which is what a summary needs, because the wording is not in the document. Extract is the closest wrong option. It returns text exactly as it appears, which suits things like a contract number or a date. Classify assigns the content to one of the categories you define, so it would give a label and not a summary. Optical character recognition reads text from images and is not a way to define how a field’s value is produced.
More guides for AI-901
- AI-901 Exam Format, Questions and Scoring
- AI-901 Study Plan: 4-Week and 8-Week
- AI-901 Prerequisites: What to Know First
- AI-901 Renewal and Retakes Explained
- Is AI-901 Worth It? An Honest Look
- AI-901 Cheat Sheet: Key Facts to Review
- AZ-900 vs AI-901: Which First?
- AB-900 vs AI-901: Which Exam First?
- AWS AIF-C01 vs AI-901: Which First?
Next step
Want more of this kind of practice? Our AI-901 practice tests are original sets with explanations for every answer option. To judge how ready you are, read is AI-901 hard?.