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 15

Question 140

A project needs the following AI behavior: A beginner app sends user text to a chat completion endpoint. What should the designer implement?

  1. Update the configured deployment name and verify compatibility with the SDK
  2. Store the approved instruction as versioned application configuration
  3. Limit context to necessary messages and apply safe summaries or retrieval
  4. Build a supported messages collection with roles and the user's content

Explanation

Correct Answer

D:

Chat APIs use structured messages rather than concatenating untyped command strings. This implementation addresses the precise goal rather than an adjacent AI task.

Incorrect Answers

A:

A model family label is not necessarily the concrete deployed endpoint. That is not sufficient here: Chat APIs use structured messages rather than concatenating untyped command strings.

B:

Versioned system messages provide a controlled behavioral baseline. That is not sufficient here: Chat APIs use structured messages rather than concatenating untyped command strings.

C:

Long irrelevant context consumes tokens and may increase latency. That is not sufficient here: Chat APIs use structured messages rather than concatenating untyped command strings.

Question 141

An Azure AI Fundamentals solution has this requirement: A service with a multilingual phone line needs a safe unsupported-language fallback. Which implementation most directly addresses it?

  1. Apply named entity recognition and preserve detected spans and categories
  2. Limit automatic language identification to supported locales and provide escalation
  3. Match the speech resource region or endpoint to the SDK configuration
  4. Escape XML special characters and validate SSML syntax

Explanation

Correct Answer

B:

Unrecognized languages must not be silently interpreted as a known one. This implementation addresses the precise goal rather than an adjacent AI task.

Incorrect Answers

A:

NER identifies named spans rather than summarizing the email. That is not sufficient here: Unrecognized languages must not be silently interpreted as a known one.

C:

Region or endpoint mismatch prevents requests reaching the correct resource. That is not sufficient here: Unrecognized languages must not be silently interpreted as a known one.

D:

SSML is XML and malformed markup breaks the synthesis request. That is not sufficient here: Unrecognized languages must not be silently interpreted as a known one.

Question 142

Troubleshooting report: The learner assumes the system performs an exact database lookup for every word. Which correction is MOST directly targeted to this failure?

  1. Review label provenance and downstream decision disparities
  2. Select a classification output for image-wide categories
  3. Demonstrate probabilistic token generation with contrasting prompts
  4. Run task-specific evaluations before choosing deployment size

Explanation

Correct Answer

C:

This correction directly addresses the reported failure. A generative language model estimates likely continuations rather than reading a stored answer paragraph. The targeted change is: Demonstrate probabilistic token generation with contrasting prompts.

Incorrect Answers

A:

Bias can enter through the targets themselves, not just inputs. In the stated case, troubleshooting report: The learner assumes the system performs an exact database lookup for every word; “Review label provenance and downstream decision disparities” does not target that failure.

B:

Classification assigns image-wide categories rather than bounding-box coordinates. In the stated case, troubleshooting report: The learner assumes the system performs an exact database lookup for every word; “Select a classification output for image-wide categories” does not target that failure.

D:

Model size alone does not establish suitability for a particular task. In the stated case, troubleshooting report: The learner assumes the system performs an exact database lookup for every word; “Run task-specific evaluations before choosing deployment size” does not target that failure.

Question 143

An AI operations team is testing two distinct capabilities. Failure (1): The application tags overall sentiment rather than entity spans. Failure (2): The system outputs a summary instead of a translation. Which TWO specific corrections should the team apply?

  1. Apply entity extraction and validate spans and categories
  2. Analyze multiple temporal segments and preserve timecodes
  3. Use translation and inspect named terms
  4. Run unsupervised clustering and evaluate cluster characteristics
  5. Use image text extraction and low-confidence human review

Explanation

Correct Answers

A:

Correct for failure (1). NER locates and categorizes named entities in text. The proposed correction directly addresses that reported defect.

C:

Correct for failure (2). Translation transfers meaning across languages rather than changing topic labels. It independently corrects the second defect.

Incorrect Answers

B:

Video understanding retains order and time, unlike single-image tagging. In this case the first outcome requires the team to use named entity recognition with entity types, and the second requires it to use machine translation with locale-specific quality review. This option does not replace those two separate controls.

D:

Clustering discovers groups based on feature similarity. In this case the first outcome requires the team to use named entity recognition with entity types, and the second requires it to use machine translation with locale-specific quality review. This option does not replace those two separate controls.

E:

Text recognition reads characters rather than merely describing visual objects. In this case the first outcome requires the team to use named entity recognition with entity types, and the second requires it to use machine translation with locale-specific quality review. This option does not replace those two separate controls.

Question 144

A solution team has selected this control: Apply named entity recognition and preserve detected spans and categories. Which observation is the most direct acceptance evidence?

  1. The selected voice reads the intended language correctly
  2. The recognizer succeeds with valid resource credentials
  3. A test email returns correct product and organization spans
  4. The converted test recording produces a faithful transcript

Explanation

Correct Answer

C:

NER identifies named spans rather than summarizing the email. The expected acceptance evidence is: A test email returns correct product and organization spans. This observation directly tests the stated behavior.

Incorrect Answers

A:

Not every voice is supported for all languages. This observation does not establish the required result: A test email returns correct product and organization spans.

B:

Audio capture success does not establish remote service authentication. This observation does not establish the required result: A test email returns correct product and organization spans.

D:

Speech recognition requires the service to decode audio correctly. This observation does not establish the required result: A test email returns correct product and organization spans.

Question 145

An Azure-based AI team must satisfy this objective: A summarizer must use a specified source document and not outside assumptions. Which design choice is best?

  1. Specify the output contract and validate returned fields against the schema
  2. Insert relevant source passages and require evidence-linked answers
  3. Use a versioned evaluation set with stable scoring and test inputs
  4. Distinguish generation variability controls from evidence-based factual accuracy

Explanation

Correct Answer

B:

Grounding constrains the answer to the available text and provenance. This implementation addresses the precise goal rather than an adjacent AI task.

Incorrect Answers

A:

Model output constraints help downstream parsing but still require validation. That is not sufficient here: Grounding constrains the answer to the available text and provenance.

C:

Anecdotal success on different prompts cannot establish an improvement. That is not sufficient here: Grounding constrains the answer to the available text and provenance.

D:

Sampling parameters influence diversity but do not verify knowledge. That is not sufficient here: Grounding constrains the answer to the available text and provenance.

Question 146

An Azure AI Fundamentals solution has this requirement: A content moderation filter blocks legitimate disability-related support requests. Which implementation most directly addresses it?

  1. Use grounded medical content, high-risk detection and appropriate expert escalation
  2. Version procedures and limit grounding to the latest authorized operational guidance
  3. Measure false positives on approved accessibility scenarios and refine thresholds
  4. Test accuracy on representative lighting and camera conditions and provide a safe fallback

Explanation

Correct Answer

C:

Safety controls should not overblock permitted assistance. This implementation addresses the precise goal rather than an adjacent AI task.

Incorrect Answers

A:

Safety mechanisms should reduce unsupported high-impact recommendations. That is not sufficient here: Safety controls should not overblock permitted assistance.

B:

A precise explanation of an obsolete procedure can still be dangerous. That is not sufficient here: Safety controls should not overblock permitted assistance.

D:

Reliability demands testing the actual operating environment, not just clean samples. That is not sufficient here: Safety controls should not overblock permitted assistance.

Question 147

An Azure-based AI team must satisfy this objective: A warehouse must count boxes visible on a pallet photograph. Which design choice is best?

  1. Use an agent orchestration plan with separate approval before booking
  2. Use an agent workflow with a permitted ticket tool and an authorization gate
  3. Use object detection and count accepted object instances
  4. Use an agent with scoped retrieval and separately authorized mutation tools

Explanation

Correct Answer

C:

Instance detections support counts where one image contains several objects. This implementation addresses the precise goal rather than an adjacent AI task.

Incorrect Answers

A:

An agent may propose actions but payment must remain permission-controlled. That is not sufficient here: Instance detections support counts where one image contains several objects.

B:

Agentic workloads combine reasoning and bounded external actions. That is not sufficient here: Instance detections support counts where one image contains several objects.

D:

Agentic workflows combine information access with controlled actions. That is not sufficient here: Instance detections support counts where one image contains several objects.

Question 148

Troubleshooting report: The team assumes matching app source code guarantees identical behavior. Which correction is MOST directly targeted to this failure?

  1. Preserve role separation and tool-side permission checks
  2. Reinforce role boundaries and enforce secrets outside prompts
  3. Reconcile dependent configuration and rerun targeted regressions
  4. Add adversarial cases and verify configured safety block and escalation behavior

Explanation

Correct Answer

C:

This correction directly addresses the reported failure. Environment drift can change behavior without code edits. The targeted change is: Reconcile dependent configuration and rerun targeted regressions.

Incorrect Answers

A:

A user cannot use text to elevate permissions beyond the application policy. In the stated case, troubleshooting report: The team assumes matching app source code guarantees identical behavior; “Preserve role separation and tool-side permission checks” does not target that failure.

B:

Developer and system constraints govern allowed operations over user requests. In the stated case, troubleshooting report: The team assumes matching app source code guarantees identical behavior; “Reinforce role boundaries and enforce secrets outside prompts” does not target that failure.

D:

Adversarial testing must measure actual configured policy enforcement rather than treating a prompt instruction as a guaranteed safety boundary. In the stated case, troubleshooting report: The team assumes matching app source code guarantees identical behavior; “Add adversarial cases and verify configured safety block and escalation behavior” does not target that failure.

Question 149

An Azure-based AI team must satisfy this objective: A visual safety check must not misidentify ambiguous objects as certain. Which design choice is best?

  1. Choose detection only if spatial object locations are required
  2. Expose uncertainty and route low-confidence image decisions to review
  3. Keep recognized visual labels separate from text-read lines and words
  4. Use optical character recognition for image text

Explanation

Correct Answer

B:

Generated visual interpretations can be wrong even when phrased fluently. This implementation addresses the precise goal rather than an adjacent AI task.

Incorrect Answers

A:

Classification predicts image-level categories while detection provides positions. That is not sufficient here: Generated visual interpretations can be wrong even when phrased fluently.

C:

Tags describe concepts while OCR reports character content. That is not sufficient here: Generated visual interpretations can be wrong even when phrased fluently.

D:

OCR extracts characters visible in pixels. That is not sufficient here: Generated visual interpretations can be wrong even when phrased fluently.

Question 150

An engineering review documents this requirement: A model describes a chart but invents values absent from the graph. Select the most appropriate solution.

  1. Choose a crop or resize preserving the requested aspect ratio and subject
  2. Ground interpretation in visible labels and qualify uncertain numerical readings
  3. Provide the actual image to a capable multimodal or OCR pipeline
  4. Detect individual package instances and count nonduplicate boxes

Explanation

Correct Answer

B:

Visual reasoning is not proof of an exact value when axes are unclear. This implementation addresses the precise goal rather than an adjacent AI task.

Incorrect Answers

A:

Smart cropping differs from simply stretching pixels. That is not sufficient here: Visual reasoning is not proof of an exact value when axes are unclear.

C:

Visual evidence is needed to read sign content. That is not sufficient here: Visual reasoning is not proof of an exact value when axes are unclear.

D:

Object counts require resolving distinct instances, not trusting one scene label. That is not sufficient here: Visual reasoning is not proof of an exact value when axes are unclear.

Microsoft Azure AI Fundamentals (AI-901) Practice Tests