Free Azure AI Fundamentals AI-901 practice questions

Sample question 1. A support team wants to find past tickets that mean the same thing as a new ticket even when they share no keywords, for example matching "cannot sign in" with "login keeps failing". The result will be a ranked list of similar tickets, not generated text. Which type of model BEST fits this requirement?

1. An image-generation model, prompted with the ticket text so that tickets producing similar pictures can be grouped together.
2. A speech-to-text model, so that tickets are converted into a normalized transcript before they are compared by exact match.
3. A chat completion model, asked to read every past ticket in each prompt and write out which ones sound similar.
4. An embeddings model, which turns each ticket into a numeric vector so that tickets with similar meaning sit close together.

Sample question 2. A developer enables guardrail annotations on a deployed model and wants the application to log, for each reply, how severe the detected violence content was and whether the guardrail actually blocked it. What does the annotation provide for the four core-harm categories?

1. A probability from 0 to 1 and the list of exact words that triggered the guardrail.
2. A single pass or fail verdict and the name of the human reviewer who approved the reply.
3. The severity level (safe, low, medium, or high) for each category, plus a filtered flag set to true or false.
4. A groundedness percentage and the citation of the public repository where matching text was found.

Sample question 3. A law firm must produce brief overviews of long vendor contracts. Reviewers insist that every sentence in an overview appear word for word in the contract, so that nothing is reworded or invented, and they want to know where in the contract each sentence came from. Which Azure Language capability fits?

1. Abstractive summarization, which composes new concise sentences that capture the main idea of the whole contract.
2. Key phrase extraction, which returns a list of the main concepts as short phrases rather than whole sentences.
3. Extractive summarization, which returns the most relevant original sentences along with their positions.
4. Sentiment analysis, which labels the overall opinion in the text.