McAfee Secure

Exam Code: AI-901

Exam Name: Microsoft Azure AI Fundamentals

Certification Provider: Microsoft

Corresponding Certification: Microsoft Certified: Azure AI Fundamentals

Microsoft AI-901 Questions & Answers

Study with Up-To-Date REAL Exam Questions and Answers from the ACTUAL Test

166 Questions & Answers with Testing Engine

"Microsoft Azure AI Fundamentals Exam", also known as AI-901 exam, is a Microsoft certification exam.

Pass your tests with the always up-to-date AI-901 Exam Engine. Your AI-901 training materials keep you at the head of the pack!

Satisfaction Guaranteed

Money Back Guarantee

Testking provides hassle-free money back guarantee with our products. That is because we have 100% trust in the abilities of our professional and experience product team, and our record is a proof of that.

99.6% PASS RATE
Was: $137.49
Now: $124.99

Microsoft AI-901 Practice Test 88

Question 870

The approved design action is: Update the configured deployment name and verify compatibility with the SDK. Which test result would BEST establish that it worked?

  1. An account-specific summary excludes cross-account conversation data
  2. The same task passes with reduced token use and no quality loss
  3. A smoke test reaches the intended new deployment and parses the output
  4. A request using the deployment name returns a response from that resource

Explanation

Correct Answer

C:

A model family label is not necessarily the concrete deployed endpoint. The expected acceptance evidence is: A smoke test reaches the intended new deployment and parses the output. This observation directly tests the stated behavior.

Incorrect Answers

A:

Source scoping enforces privacy and also improves answer relevance. This observation does not establish the required result: A smoke test reaches the intended new deployment and parses the output.

B:

Long irrelevant context consumes tokens and may increase latency. This observation does not establish the required result: A smoke test reaches the intended new deployment and parses the output.

D:

An inferencing request is routed to a concrete deployment configured in the resource. This observation does not establish the required result: A smoke test reaches the intended new deployment and parses the output.

Question 871

An AI operations team is testing two distinct capabilities. Failure (1): The system invents tasks absent from any transcript segment. Failure (2): The interface says stopped while background capture continues. Which TWO specific corrections should the team apply?

  1. Enable supported diarization and validate turn assignments
  2. Require source-attributed action extraction
  3. Send valid audio to a speech recognition endpoint
  4. Close the capture stream and update permission controls
  5. Provide valid SSML break elements to the synthesis service

Explanation

Correct Answers

B:

Correct for failure (1). Transcription obtains words; summarization extracts decisions and tasks. The proposed correction directly addresses that reported defect.

D:

Correct for failure (2). User control of capture must be enforced at the device or application layer. It independently corrects the second defect.

Incorrect Answers

A:

Speaker diarization separates turns and identities at a conversation level. The stated pair of failures instead requires “Require source-attributed action extraction” and “Close the capture stream and update permission controls”; this option cannot substitute for both.

C:

Speech recognition converts waveform features to text tokens. It does not complete the two decisions required in this question.

E:

SSML controls aspects of prosody and timing for synthesized speech. The stated pair of failures instead requires “Require source-attributed action extraction” and “Close the capture stream and update permission controls”; this option cannot substitute for both.

Question 872

Troubleshooting report: The assistant confidently calculates a dose from unknown body weight. Which correction is MOST directly targeted to this failure?

  1. Reject the incomplete input and request professional confirmation
  2. Audit and correct all cross-region ancillary data flows
  3. Remove protected payloads from logging and enforce redaction tests
  4. Add usable transcript and caption output

Explanation

Correct Answer

A:

This correction directly addresses the reported failure. Safety-critical recommendations require verified constraints outside open-ended generation. The targeted change is: Reject the incomplete input and request professional confirmation.

Incorrect Answers

B:

Privacy and compliance require controlling the full data path. In the stated case, troubleshooting report: The assistant confidently calculates a dose from unknown body weight; “Audit and correct all cross-region ancillary data flows” does not target that failure.

C:

Troubleshooting does not justify reproducing sensitive application records. In the stated case, troubleshooting report: The assistant confidently calculates a dose from unknown body weight; “Remove protected payloads from logging and enforce redaction tests” does not target that failure.

D:

Accessible information must not depend on hearing audio. In the stated case, troubleshooting report: The assistant confidently calculates a dose from unknown body weight; “Add usable transcript and caption output” does not target that failure.

Question 873

An AI operations team is testing two distinct capabilities. Failure (1): The team attempts fine-tuning each time a price changes. Failure (2): The team always selects the largest model without measuring task fit. Which TWO specific corrections should the team apply?

  1. Set the API request to the correct deployment name and endpoint
  2. Run task-specific evaluations before choosing deployment size
  3. Evaluate minority-class recall and threshold tradeoffs
  4. Chunk and retrieve relevant source sections before inference
  5. Refresh indexed grounding sources and verify their effective date

Explanation

Correct Answers

B:

Correct for failure (2). Model size alone does not establish suitability for a particular task. It independently corrects the second defect.

E:

Correct for failure (1). External retrieval updates factual context without retraining model weights. The proposed correction directly addresses that reported defect.

Incorrect Answers

A:

Applications call a deployed model resource rather than a generic marketing model name. The stated pair of failures instead requires “Run task-specific evaluations before choosing deployment size” and “Refresh indexed grounding sources and verify their effective date”; this option cannot substitute for both.

C:

High accuracy can conceal failure to detect rare positive events. The stated pair of failures instead requires “Run task-specific evaluations before choosing deployment size” and “Refresh indexed grounding sources and verify their effective date”; this option cannot substitute for both.

D:

A finite context window limits how much material can be processed at once. The stated pair of failures instead requires “Run task-specific evaluations before choosing deployment size” and “Refresh indexed grounding sources and verify their effective date”; this option cannot substitute for both.

Question 874

An engineering review documents this requirement: A model output is deterministic during testing but changes after the deployment version is updated. Select the most appropriate solution.

  1. Benchmark both on representative quality, latency and cost thresholds
  2. Adjust supported sampling parameters while preserving safety and quality tests
  3. Track deployment versions, sampling parameters and test inputs together
  4. Use an image-classification model with class probabilities

Explanation

Correct Answer

C:

Reproducibility depends on more than the visible prompt text. This implementation addresses the precise goal rather than an adjacent AI task.

Incorrect Answers

A:

Model size alone does not establish suitability for a particular task. That is not sufficient here: Reproducibility depends on more than the visible prompt text.

B:

More varied token sampling can increase creative alternatives. That is not sufficient here: Reproducibility depends on more than the visible prompt text.

D:

Classification assigns image-wide categories rather than bounding-box coordinates. That is not sufficient here: Reproducibility depends on more than the visible prompt text.

Question 875

An Azure AI Fundamentals solution has this requirement: An agent has several similar tools for customer lookup and refund creation. Which implementation most directly addresses it?

  1. Externalize endpoint and deployment configuration per environment
  2. Give each tool a discriminating description and narrow permissions
  3. Implement a network timeout/retry strategy that avoids duplicate tool effects
  4. Use representative test prompts and inspect execution and tool traces

Explanation

Correct Answer

B:

Clear tool contracts reduce confusion between read and write operations. This implementation addresses the precise goal rather than an adjacent AI task.

Incorrect Answers

A:

A correct code path can still target the wrong resource after migration. That is not sufficient here: Clear tool contracts reduce confusion between read and write operations.

C:

Client retries should distinguish read requests from side-effecting tool calls. That is not sufficient here: Clear tool contracts reduce confusion between read and write operations.

D:

An answer that sounds plausible may conceal failed tool calls. That is not sufficient here: Clear tool contracts reduce confusion between read and write operations.

Question 876

A solution team has selected this control: Combine audio transcription, speaker segmentation and grounded summarization. Which observation is the most direct acceptance evidence?

  1. The product shape remains unchanged against the new background
  2. The minutes correctly associate tasks with speakers and source timestamps
  3. The transcript contains the actual spoken issue and timestamp
  4. Totals, dates and merchant names match receipt evidence

Explanation

Correct Answer

B:

Attribution distinguishes who committed to each action. The expected acceptance evidence is: The minutes correctly associate tasks with speakers and source timestamps. This observation directly tests the stated behavior.

Incorrect Answers

A:

Editing changes selected pixels while leaving product regions intact. This observation does not establish the required result: The minutes correctly associate tasks with speakers and source timestamps.

C:

Speech recognition converts speech into words suitable for search. This observation does not establish the required result: The minutes correctly associate tasks with speakers and source timestamps.

D:

Images of receipts need OCR/layout extraction before typed business values. This observation does not establish the required result: The minutes correctly associate tasks with speakers and source timestamps.

Question 877

The approved design action is: Explain that tokens are model input/output units, not necessarily entire words. Which test result would BEST establish that it worked?

  1. The output is produced as a sequence of tokens conditioned on the prompt
  2. A permitted larger output budget allows the complete response
  3. Repeated structured cases follow the expected schema more consistently
  4. A tokenization example shows one word can occupy multiple tokens

Explanation

Correct Answer

D:

Tokenization splits text into subword or other units for model processing. The expected acceptance evidence is: A tokenization example shows one word can occupy multiple tokens. This observation directly tests the stated behavior.

Incorrect Answers

A:

A generative language model estimates likely continuations rather than reading a stored answer paragraph. This observation does not establish the required result: A tokenization example shows one word can occupy multiple tokens.

B:

Generation stops when an output budget or stop condition is reached. This observation does not establish the required result: A tokenization example shows one word can occupy multiple tokens.

C:

Lower randomness often improves consistency but does not guarantee correctness. This observation does not establish the required result: A tokenization example shows one word can occupy multiple tokens.

Question 878

Troubleshooting report: The review tests only average accuracy on the majority group. Which correction is MOST directly targeted to this failure?

  1. Rebalance safety tests and adjust erroneous dialect associations
  2. Replace impersonating presentation with explicit synthetic-agent notice
  3. Add subgroup and matched-pair fairness tests before approval
  4. Add independent risk review and verifiable test artifacts

Explanation

Correct Answer

C:

This correction directly addresses the reported failure. Fairness checks compare materially equivalent cases rather than raw group averages alone. The targeted change is: Add subgroup and matched-pair fairness tests before approval.

Incorrect Answers

A:

Fairness requires discriminating actual harm from linguistic variety. In the stated case, troubleshooting report: The review tests only average accuracy on the majority group; “Rebalance safety tests and adjust erroneous dialect associations” does not target that failure.

B:

Transparency prevents users from falsely assuming they are speaking with a human. In the stated case, troubleshooting report: The review tests only average accuracy on the majority group; “Replace impersonating presentation with explicit synthetic-agent notice” does not target that failure.

D:

Self-attestation is not independent proof of regulatory controls. In the stated case, troubleshooting report: The review tests only average accuracy on the majority group; “Add independent risk review and verifiable test artifacts” does not target that failure.

Question 879

A project needs the following AI behavior: A designer asks for several visually distinct variations of an original concept. What should the designer implement?

  1. Detect individual package instances and count nonduplicate boxes
  2. Use image classification for the requested whole-image class
  3. Generate multiple authorized image candidates and assess against the brief
  4. Request image tagging for the selected visual features

Explanation

Correct Answer

C:

Image-generation variability can be useful but does not guarantee suitability. This implementation addresses the precise goal rather than an adjacent AI task.

Incorrect Answers

A:

Object counts require resolving distinct instances, not trusting one scene label. That is not sufficient here: Image-generation variability can be useful but does not guarantee suitability.

B:

Image-level classification is distinct from object localization. That is not sufficient here: Image-generation variability can be useful but does not guarantee suitability.

D:

Tags summarize recognized concepts as labels. That is not sufficient here: Image-generation variability can be useful but does not guarantee suitability.

Question 880

Troubleshooting report: The tool performs sentiment classification instead of translation. Which correction is MOST directly targeted to this failure?

  1. Pass supported audio content to an audio-capable model
  2. Require source-attributed action extraction
  3. Select the correct source/target language translation path
  4. Load the deployed speech resource's region and endpoint

Explanation

Correct Answer

C:

This correction directly addresses the reported failure. Translation requires semantic equivalence and language direction. The targeted change is: Select the correct source/target language translation path.

Incorrect Answers

A:

A text-only request cannot interpret sound content by file name alone. In the stated case, troubleshooting report: The tool performs sentiment classification instead of translation; “Pass supported audio content to an audio-capable model” does not target that failure.

B:

Transcription obtains words; summarization extracts decisions and tasks. In the stated case, troubleshooting report: The tool performs sentiment classification instead of translation; “Require source-attributed action extraction” does not target that failure.

D:

Region or endpoint mismatch prevents requests reaching the correct resource. In the stated case, troubleshooting report: The tool performs sentiment classification instead of translation; “Load the deployed speech resource's region and endpoint” does not target that failure.

Microsoft Azure AI Fundamentals (AI-901) Practice Tests