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 84

Question 830

An Azure AI Fundamentals solution has this requirement: A support app uses a tool that should only list orders. Which implementation most directly addresses it?

  1. Expose a read-only tool contract rather than a broad administrative API
  2. Preserve the higher-priority application constraints in the message hierarchy
  3. Use operation IDs and read-back receipts before replaying external side effects
  4. Honor service throttling and use bounded exponential backoff

Explanation

Correct Answer

A:

Least-privilege tool methods reduce accidental external actions. This implementation addresses the precise goal rather than an adjacent AI task.

Incorrect Answers

B:

Developer and system constraints govern allowed operations over user requests. That is not sufficient here: Least-privilege tool methods reduce accidental external actions.

C:

A timeout does not prove that the original mutation failed. That is not sufficient here: Least-privilege tool methods reduce accidental external actions.

D:

Rate limits protect the service and uncontrolled retries amplify overload. That is not sufficient here: Least-privilege tool methods reduce accidental external actions.

Question 831

An Azure AI solution uses the following approved control: Check the selected credential, identity scope and target resource permission. What observable result should pass QA?

  1. The difference can be tied to an explicit configuration mismatch
  2. The service processes role-tagged messages with the expected application behavior
  3. The same task passes with reduced token use and no quality loss
  4. A valid identity with required resource role succeeds in the test

Explanation

Correct Answer

D:

Authentication proves caller identity and authorization grants resource access. The expected acceptance evidence is: A valid identity with required resource role succeeds in the test. This observation directly tests the stated behavior.

Incorrect Answers

A:

Environment drift can change behavior without code edits. This observation does not establish the required result: A valid identity with required resource role succeeds in the test.

B:

Chat APIs use structured messages rather than concatenating untyped command strings. This observation does not establish the required result: A valid identity with required resource role succeeds in the test.

C:

Long irrelevant context consumes tokens and may increase latency. This observation does not establish the required result: A valid identity with required resource role succeeds in the test.

Question 832

A project needs the following AI behavior: An insurance processor needs handwritten notes and checked boxes from photographed claims forms. What should the designer implement?

  1. Use layout-aware extraction and selection-mark analysis
  2. Use sentiment analysis to estimate positive, negative, or neutral tone
  3. Use object detection and count accepted object instances
  4. Use image segmentation to map visual damage areas instead of assigning one global defect severity

Explanation

Correct Answer

A:

Different form elements require more than ordinary running-text OCR. This implementation addresses the precise goal rather than an adjacent AI task.

Incorrect Answers

B:

Sentiment analysis infers expressed attitude rather than extracting names. That is not sufficient here: Different form elements require more than ordinary running-text OCR.

C:

Instance detections support counts where one image contains several objects. That is not sufficient here: Different form elements require more than ordinary running-text OCR.

D:

Pixel-level region masks localize building damage more precisely than an overall image class label. That is not sufficient here: Different form elements require more than ordinary running-text OCR.

Question 833

An AI operations team is testing two distinct capabilities. Failure (1): The completion text invents a search result from model memory. Failure (2): The prototype launches unnecessary specialist agents for one lookup. Which TWO specific corrections should the team apply?

  1. Create and test a simple one-agent tool workflow
  2. Require real tool calls and distinguish pending from completed
  3. Reconcile dependent configuration and rerun targeted regressions
  4. Attach idempotency keys and verify tool receipts after reconnecting
  5. Use an abstention or clarification branch when no source matches

Explanation

Correct Answers

A:

Correct for failure (2). A single agent can orchestrate a bounded tool without a multi-agent system. It independently corrects the second defect.

B:

Correct for failure (1). Generated text does not prove external execution. The proposed correction directly addresses that reported defect.

Incorrect Answers

C:

Environment drift can change behavior without code edits. It does not complete the two decisions required in this question.

D:

Client retries should distinguish read requests from side-effecting tool calls. The stated pair of failures instead requires “Create and test a simple one-agent tool workflow” and “Require real tool calls and distinguish pending from completed”; this option cannot substitute for both.

E:

An empty result is not positive evidence for a generated factual claim. The stated pair of failures instead requires “Create and test a simple one-agent tool workflow” and “Require real tool calls and distinguish pending from completed”; this option cannot substitute for both.

Question 834

The approved design action is: Generate and edit concise factual alt text describing significant content. Which test result would BEST establish that it worked?

  1. The response distinguishes readable values from estimates
  2. The crop keeps the subject in view at the specified aspect ratio
  3. A nonvisual test user understands the relevant scene content
  4. The returned string matches the labeled serial number

Explanation

Correct Answer

C:

Accessible captions should communicate image purpose without invented details. The expected acceptance evidence is: A nonvisual test user understands the relevant scene content. This observation directly tests the stated behavior.

Incorrect Answers

A:

Visual reasoning is not proof of an exact value when axes are unclear. This observation does not establish the required result: A nonvisual test user understands the relevant scene content.

B:

Smart cropping attempts to retain salient regions within target dimensions. This observation does not establish the required result: A nonvisual test user understands the relevant scene content.

D:

OCR extracts characters visible in pixels. This observation does not establish the required result: A nonvisual test user understands the relevant scene content.

Question 835

An acceptance tester needs to validate: Use image segmentation to map visual damage areas instead of assigning one global defect severity. Which observed result proves the intended behavior?

  1. A consented test delivers equivalent meaning in the listener's language
  2. The minutes correctly associate tasks with speakers and source timestamps
  3. Parsed fields match labeled form values and missing fields remain explicit
  4. A labeled photograph produces a correctly aligned damage-region mask rather than a global category

Explanation

Correct Answer

D:

Pixel-level region masks localize building damage more precisely than an overall image class label. The expected acceptance evidence is: A labeled photograph produces a correctly aligned damage-region mask rather than a global category. This observation directly tests the stated behavior.

Incorrect Answers

A:

Speech translation spans audio input, semantic transfer and optional output audio. This observation does not establish the required result: A labeled photograph produces a correctly aligned damage-region mask rather than a global category.

B:

Attribution distinguishes who committed to each action. This observation does not establish the required result: A labeled photograph produces a correctly aligned damage-region mask rather than a global category.

C:

Document extraction maps unstructured layouts into structured values. This observation does not establish the required result: A labeled photograph produces a correctly aligned damage-region mask rather than a global category.

Question 836

A project needs the following AI behavior: A factory safety assistant gives advice with a stale equipment procedure. What should the designer implement?

  1. Evaluate safety again after tool and retrieval results are incorporated
  2. Redact personal fields before telemetry storage and control log access
  3. Version procedures and limit grounding to the latest authorized operational guidance
  4. Use validated data retrieval and source-backed numerical responses

Explanation

Correct Answer

C:

A precise explanation of an obsolete procedure can still be dangerous. This implementation addresses the precise goal rather than an adjacent AI task.

Incorrect Answers

A:

Threats can enter later stages even when the original user request was harmless. That is not sufficient here: A precise explanation of an obsolete procedure can still be dangerous.

B:

Troubleshooting does not justify reproducing sensitive application records. That is not sufficient here: A precise explanation of an obsolete procedure can still be dangerous.

D:

A numeric-looking answer is not proof that a measure was computed. That is not sufficient here: A precise explanation of an obsolete procedure can still be dangerous.

Question 837

The approved design action is: Insert relevant source passages and require evidence-linked answers. Which test result would BEST establish that it worked?

  1. The same app points to the intended test or production resource by configuration
  2. The agent retrieves an authorized relevant passage and cites it
  3. The summary references statements supported by the supplied passage
  4. Browser traffic cannot access the protected model credential

Explanation

Correct Answer

C:

Grounding constrains the answer to the available text and provenance. The expected acceptance evidence is: The summary references statements supported by the supplied passage. This observation directly tests the stated behavior.

Incorrect Answers

A:

A correct code path can still target the wrong resource after migration. This observation does not establish the required result: The summary references statements supported by the supplied passage.

B:

Tool descriptions guide selection, while backend rules enforce allowed sources. This observation does not establish the required result: The summary references statements supported by the supplied passage.

D:

Backend controls protect credentials, quotas and data policy. This observation does not establish the required result: The summary references statements supported by the supplied passage.

Question 838

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

  1. The same facts are consistently presented in the required style
  2. Held-out defect images meet the tested localization and per-class detection thresholds
  3. A regression test records model version and key generation settings
  4. The model/deployment capability check and a spoken-audio test confirm that this endpoint supports transcription

Explanation

Correct Answer

B:

An appropriate spatial detection model must localize the defect classes, not merely assign an overall picture label. The expected acceptance evidence is: Held-out defect images meet the tested localization and per-class detection thresholds. This observation directly tests the stated behavior.

Incorrect Answers

A:

Clear prompting can control tone without model retraining. This observation does not establish the required result: Held-out defect images meet the tested localization and per-class detection thresholds.

C:

Reproducibility depends on more than the visible prompt text. This observation does not establish the required result: Held-out defect images meet the tested localization and per-class detection thresholds.

D:

Model capabilities must include the required audio modality before the client can obtain speech-to-text results. This observation does not establish the required result: Held-out defect images meet the tested localization and per-class detection thresholds.

Question 839

Troubleshooting report: The extractor reports a signed agreement based solely on a typed name. Which correction is MOST directly targeted to this failure?

  1. Verify the actual source signature region or use explicitly approved preview detection, never infer signing from a typed name
  2. Enable supported document confidence and evaluate thresholds against labeled results and business rules
  3. Use layout-aware chunking and source identifiers
  4. Preserve actual section/page references through retrieval

Explanation

Correct Answer

A:

This correction directly addresses the reported failure. Selection-mark extraction is available with GA document analysis, but dedicated signature-location detection is available only in the 2026-06-01-preview API and cannot establish legal validity. The targeted change is: Verify the actual source signature region or use explicitly approved preview detection, never infer signing from a typed name.

Incorrect Answers

B:

Content Understanding confidence applies to supported document analyzers; confidence estimates alone do not prove field correctness. In the stated case, troubleshooting report: The extractor reports a signed agreement based solely on a typed name; “Enable supported document confidence and evaluate thresholds against labeled results and business rules” does not target that failure.

C:

RAG preparation should preserve document structure and source references. In the stated case, troubleshooting report: The extractor reports a signed agreement based solely on a typed name; “Use layout-aware chunking and source identifiers” does not target that failure.

D:

Structured layout and provenance let users verify a quoted clause. In the stated case, troubleshooting report: The extractor reports a signed agreement based solely on a typed name; “Preserve actual section/page references through retrieval” does not target that failure.

Question 840

An engineering review documents this requirement: A marketer wants the same generated illustration to follow a specific brand palette. Select the most appropriate solution.

  1. Use region-aware captioning or a cropped image focused on that location
  2. Generate multiple authorized image candidates and assess against the brief
  3. Choose detection only if spatial object locations are required
  4. Include palette and composition constraints in the image prompt and review output

Explanation

Correct Answer

D:

Explicit visual guidance can influence generated appearance but is not absolute. This implementation addresses the precise goal rather than an adjacent AI task.

Incorrect Answers

A:

Specific regions can carry details missed by a whole-image caption. That is not sufficient here: Explicit visual guidance can influence generated appearance but is not absolute.

B:

Image-generation variability can be useful but does not guarantee suitability. That is not sufficient here: Explicit visual guidance can influence generated appearance but is not absolute.

C:

Classification predicts image-level categories while detection provides positions. That is not sufficient here: Explicit visual guidance can influence generated appearance but is not absolute.

Microsoft Azure AI Fundamentals (AI-901) Practice Tests