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 19

Question 180

Troubleshooting report: Action items are generated with no corresponding utterance. Which correction is MOST directly targeted to this failure?

  1. Mark the field unreadable and request manual correction
  2. Keep transcript anchors and validate extracted tasks
  3. Preserve segment timestamps and review rapid events separately when sampling might miss them
  4. Run a supported OCR or visual text extraction feature

Explanation

Correct Answer

B:

This correction directly addresses the reported failure. Transcription and summarization are separate results with different validation. The targeted change is: Keep transcript anchors and validate extracted tasks.

Incorrect Answers

A:

Low-quality inputs can make extraction genuinely uncertain. In the stated case, troubleshooting report: Action items are generated with no corresponding utterance; “Mark the field unreadable and request manual correction” does not target that failure.

C:

Temporal evidence supports the timing of observed events, but approximately one frame per second cannot guarantee subsecond event capture. In the stated case, troubleshooting report: Action items are generated with no corresponding utterance; “Preserve segment timestamps and review rapid events separately when sampling might miss them” does not target that failure.

D:

OCR is appropriate for characters embedded in a photograph. In the stated case, troubleshooting report: Action items are generated with no corresponding utterance; “Run a supported OCR or visual text extraction feature” does not target that failure.

Question 181

A solution team has selected this control: Distinguish no speech detected from recognized verbal denial. Which observation is the most direct acceptance evidence?

  1. The transcript matches a labeled spoken test phrase
  2. A request succeeds after using the exact resource region
  3. The narration pauses at marked section boundaries
  4. Silent input yields a repeat or timeout response rather than a denial

Explanation

Correct Answer

D:

An empty recognition result is not semantically equivalent to the word no. The expected acceptance evidence is: Silent input yields a repeat or timeout response rather than a denial. This observation directly tests the stated behavior.

Incorrect Answers

A:

Speech recognition converts waveform features to text tokens. This observation does not establish the required result: Silent input yields a repeat or timeout response rather than a denial.

B:

Region or endpoint mismatch prevents requests reaching the correct resource. This observation does not establish the required result: Silent input yields a repeat or timeout response rather than a denial.

C:

SSML controls aspects of prosody and timing for synthesized speech. This observation does not establish the required result: Silent input yields a repeat or timeout response rather than a denial.

Question 182

A project needs the following AI behavior: A student asks what a token is in generative text. What should the designer implement?

  1. Explain that tokens are model input/output units, not necessarily entire words
  2. Use a supervised classification model with representative labeled cases
  3. Explain that model responses depend on the prompt and available context window
  4. Check output-token limits and context-budget allocation

Explanation

Correct Answer

A:

Tokenization splits text into subword or other units for model processing. This implementation addresses the precise goal rather than an adjacent AI task.

Incorrect Answers

B:

Classification predicts discrete labels rather than a continuous numeric target. That is not sufficient here: Tokenization splits text into subword or other units for model processing.

C:

A model conditions its output on supplied instructions and conversation text. That is not sufficient here: Tokenization splits text into subword or other units for model processing.

D:

Generation stops when an output budget or stop condition is reached. That is not sufficient here: Tokenization splits text into subword or other units for model processing.

Question 183

An acceptance tester needs to validate: Choose a model size using representative quality and cost benchmarks. Which observed result proves the intended behavior?

  1. A trace shows the task finished and its output is available
  2. The chosen model meets labeled performance targets at lower cost
  3. A request using the deployment name returns a response from that resource
  4. The agent retrieves an authorized relevant passage and cites it

Explanation

Correct Answer

B:

Lower-cost deployment is acceptable only if accuracy and safety floors remain met. The expected acceptance evidence is: The chosen model meets labeled performance targets at lower cost. This observation directly tests the stated behavior.

Incorrect Answers

A:

Submission is not the same as successful completion. This observation does not establish the required result: The chosen model meets labeled performance targets at lower cost.

C:

An inferencing request is routed to a concrete deployment configured in the resource. This observation does not establish the required result: The chosen model meets labeled performance targets at lower cost.

D:

Tool descriptions guide selection, while backend rules enforce allowed sources. This observation does not establish the required result: The chosen model meets labeled performance targets at lower cost.

Question 184

A project needs the following AI behavior: A support app combines sentiment analysis and extraction on one message. What should the designer implement?

  1. Select a supported synthesis voice ID rather than altering transcript text
  2. Run the required text analysis operations separately and preserve distinct result types
  3. Evaluate word error rate on the same held-out audio samples
  4. Use a supported diarization or speaker-labeling capability where available

Explanation

Correct Answer

B:

Sentiment polarity and named entity spans answer different questions. This implementation addresses the precise goal rather than an adjacent AI task.

Incorrect Answers

A:

Voice timbre is a synthesis configuration, not speech recognition output. That is not sufficient here: Sentiment polarity and named entity spans answer different questions.

C:

Comparable labeled audio prevents misleading quality claims. That is not sufficient here: Sentiment polarity and named entity spans answer different questions.

D:

Speaker diarization separates turns and identities at a conversation level. That is not sufficient here: Sentiment polarity and named entity spans answer different questions.

Question 185

An AI operations team is testing two distinct capabilities. Failure (1): The summarizer retrieves all tenant chats by broad keyword search. Failure (2): Large requests fail due to unbounded prompt content. Which TWO specific corrections should the team apply?

  1. Create and test a simple one-agent tool workflow
  2. Apply case and role filters before providing model context
  3. Freeze representative evaluation cases and compare versions
  4. Constrain request size and trim or chunk input before sending
  5. Add approval state and verified role checks

Explanation

Correct Answers

B:

Correct for failure (1). Source scoping enforces privacy and also improves answer relevance. The proposed correction directly addresses that reported defect.

D:

Correct for failure (2). Capabilities and limits vary by model deployment. It independently corrects the second defect.

Incorrect Answers

A:

A single agent can orchestrate a bounded tool without a multi-agent system. In this case the first outcome requires the team to limit retrieval and chat context to the authenticated case scope, and the second requires it to check supported context and output limits for the chosen deployment. This option does not replace those two separate controls.

C:

Anecdotal success on different prompts cannot establish an improvement. In this case the first outcome requires the team to limit retrieval and chat context to the authenticated case scope, and the second requires it to check supported context and output limits for the chosen deployment. This option does not replace those two separate controls.

E:

High-impact side effects need permissions independent of conversational language. In this case the first outcome requires the team to limit retrieval and chat context to the authenticated case scope, and the second requires it to check supported context and output limits for the chosen deployment. This option does not replace those two separate controls.

Question 186

A project needs the following AI behavior: A transcript must include timing to synchronize captions with recorded audio. What should the designer implement?

  1. Use Speech Synthesis Markup Language pauses or breaks where supported
  2. Limit automatic language identification to supported locales and provide escalation
  3. Distinguish no speech detected from recognized verbal denial
  4. Use word or utterance timestamps from a supported recognition output

Explanation

Correct Answer

D:

Timecodes align text segments to audio playback. This implementation addresses the precise goal rather than an adjacent AI task.

Incorrect Answers

A:

SSML controls aspects of prosody and timing for synthesized speech. That is not sufficient here: Timecodes align text segments to audio playback.

B:

Unrecognized languages must not be silently interpreted as a known one. That is not sufficient here: Timecodes align text segments to audio playback.

C:

An empty recognition result is not semantically equivalent to the word no. That is not sufficient here: Timecodes align text segments to audio playback.

Question 187

An Azure AI Fundamentals solution has this requirement: An audiobook reader must choose a supported voice for a particular locale. Which implementation most directly addresses it?

  1. Call a supported text sentiment-analysis feature and interpret returned polarity
  2. Check synthesized voice availability for the selected locale and region
  3. Use supported SSML phoneme or say-as controls for pronunciation
  4. Use a supported diarization or speaker-labeling capability where available

Explanation

Correct Answer

B:

Not every voice is supported for all languages. This implementation addresses the precise goal rather than an adjacent AI task.

Incorrect Answers

A:

Sentiment output classifies attitudes expressed in the submitted text. That is not sufficient here: Not every voice is supported for all languages.

C:

SSML can guide pronunciation beyond plain-text guessing. That is not sufficient here: Not every voice is supported for all languages.

D:

Speaker diarization separates turns and identities at a conversation level. That is not sufficient here: Not every voice is supported for all languages.

Question 188

An AI operations team is testing two distinct capabilities. Failure (1): Only total monthly bill is visible with no per-call metrics. Failure (2): The team assumes English accuracy proves all language support. Which TWO specific corrections should the team apply?

  1. Rewrite the prompt with explicit acceptance criteria
  2. Run locale-specific evaluations and fallback for unsupported cases
  3. Apply case and role filters before providing model context
  4. Capture per-request usage and attribute it to versioned deployments
  5. Construct messages in the documented SDK request format

Explanation

Correct Answers

B:

Correct for failure (2). Language coverage can vary by model and deployment. It independently corrects the second defect.

D:

Correct for failure (1). Usage telemetry reveals context-growth and unexpected repetition. The proposed correction directly addresses that reported defect.

Incorrect Answers

A:

Specific instructions reduce ambiguity about what constitutes success. In this case the first outcome requires the team to record input/output token usage by model and prompt version, and the second requires it to choose a supported multilingual model or preprocessing route and test each locale. This option does not replace those two separate controls.

C:

Source scoping enforces privacy and also improves answer relevance. In this case the first outcome requires the team to record input/output token usage by model and prompt version, and the second requires it to choose a supported multilingual model or preprocessing route and test each locale. This option does not replace those two separate controls.

E:

Chat APIs use structured messages rather than concatenating untyped command strings. In this case the first outcome requires the team to record input/output token usage by model and prompt version, and the second requires it to choose a supported multilingual model or preprocessing route and test each locale. This option does not replace those two separate controls.

Question 189

Troubleshooting report: Only an audio response is available for critical information. Which correction is MOST directly targeted to this failure?

  1. Add independent risk review and verifiable test artifacts
  2. Label and evaluate representative scans by acquisition equipment
  3. Add usable transcript and caption output
  4. Remove the false attribution and require documentary consent

Explanation

Correct Answer

C:

This correction directly addresses the reported failure. Accessible information must not depend on hearing audio. The targeted change is: Add usable transcript and caption output.

Incorrect Answers

A:

Self-attestation is not independent proof of regulatory controls. In the stated case, troubleshooting report: Only an audio response is available for critical information; “Add independent risk review and verifiable test artifacts” does not target that failure.

B:

Reliability requires detecting distribution shifts hidden in aggregate accuracy. In the stated case, troubleshooting report: Only an audio response is available for critical information; “Label and evaluate representative scans by acquisition equipment” does not target that failure.

D:

Transparency prohibits presenting invented testimony as genuine. In the stated case, troubleshooting report: Only an audio response is available for critical information; “Remove the false attribution and require documentary consent” does not target that failure.

Question 190

The approved design action is: Use object detection that returns labels and bounding boxes. Which test result would BEST establish that it worked?

  1. A correctly converted test image processes successfully
  2. Expected objects have correctly positioned boxes
  3. The thumbnail preserves subject proportions and important content
  4. The answer reproduces the visible sign text accurately

Explanation

Correct Answer

B:

Object detection reports which objects appear and where they occur. The expected acceptance evidence is: Expected objects have correctly positioned boxes. This observation directly tests the stated behavior.

Incorrect Answers

A:

Vision endpoints impose supported input formats and limits. This observation does not establish the required result: Expected objects have correctly positioned boxes.

C:

Smart cropping differs from simply stretching pixels. This observation does not establish the required result: Expected objects have correctly positioned boxes.

D:

Visual evidence is needed to read sign content. This observation does not establish the required result: Expected objects have correctly positioned boxes.

Microsoft Azure AI Fundamentals (AI-901) Practice Tests