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 35

Question 340

An engineering review documents this requirement: A fleet operator wants to predict whether a vehicle will fail in the next week. Select the most appropriate solution.

  1. Use retrieval-augmented generation with actual source citations
  2. Use a visual classification or detection model appropriate to whether location matters
  3. Use supervised classification with labeled maintenance outcomes
  4. Use sentiment analysis to estimate positive, negative, or neutral tone

Explanation

Correct Answer

C:

A failure probability or class is a discrete decision under a defined horizon. This implementation addresses the precise goal rather than an adjacent AI task.

Incorrect Answers

A:

RAG retrieves current external evidence for a generated response. That is not sufficient here: A failure probability or class is a discrete decision under a defined horizon.

B:

The task chooses a label, with localization needed only when multiple signs appear. That is not sufficient here: A failure probability or class is a discrete decision under a defined horizon.

D:

Sentiment analysis infers expressed attitude rather than extracting names. That is not sufficient here: A failure probability or class is a discrete decision under a defined horizon.

Question 341

An Azure AI Fundamentals solution has this requirement: A team wants a single-agent prototype that can use a weather or lookup tool. Which implementation most directly addresses it?

  1. Use managed identity or protected secret storage rather than embedding keys in code
  2. Create one agent in the Foundry portal with scoped tool definitions
  3. Use operation IDs and read-back receipts before replaying external side effects
  4. Use tenant-specific identity, endpoint configuration and permissions

Explanation

Correct Answer

B:

A single agent can orchestrate a bounded tool without a multi-agent system. This implementation addresses the precise goal rather than an adjacent AI task.

Incorrect Answers

A:

Credentials should not appear in repositories, prompt logs or client binaries. That is not sufficient here: A single agent can orchestrate a bounded tool without a multi-agent system.

C:

A timeout does not prove that the original mutation failed. That is not sufficient here: A single agent can orchestrate a bounded tool without a multi-agent system.

D:

Copying code does not grant access to another tenant's resources. That is not sufficient here: A single agent can orchestrate a bounded tool without a multi-agent system.

Question 342

An Azure-based AI team must satisfy this objective: A shop wants a clear description of an object located in the upper-left of a photo. Which design choice is best?

  1. Use a supported dense-captioning capability where available
  2. Use region-aware captioning or a cropped image focused on that location
  3. Preserve region coordinates with extracted text or field evidence
  4. Provide the actual image to a capable multimodal or OCR pipeline

Explanation

Correct Answer

B:

Specific regions can carry details missed by a whole-image caption. This implementation addresses the precise goal rather than an adjacent AI task.

Incorrect Answers

A:

Dense captions describe localized regions beyond a single global caption. That is not sufficient here: Specific regions can carry details missed by a whole-image caption.

C:

Spatial provenance supports later verification of recognition results. That is not sufficient here: Specific regions can carry details missed by a whole-image caption.

D:

Visual evidence is needed to read sign content. That is not sufficient here: Specific regions can carry details missed by a whole-image caption.

Question 343

An engineering review documents this requirement: A reviewer needs to distinguish image classification from object detection. Select the most appropriate solution.

  1. Use evaluated confidence thresholds aligned to the business error cost
  2. Use object detection that returns labels and bounding boxes
  3. Choose detection only if spatial object locations are required
  4. Use region-aware captioning or a cropped image focused on that location

Explanation

Correct Answer

C:

Classification predicts image-level categories while detection provides positions. This implementation addresses the precise goal rather than an adjacent AI task.

Incorrect Answers

A:

Confidence scores require calibration and contextual acceptance policy. That is not sufficient here: Classification predicts image-level categories while detection provides positions.

B:

Object detection reports which objects appear and where they occur. That is not sufficient here: Classification predicts image-level categories while detection provides positions.

D:

Specific regions can carry details missed by a whole-image caption. That is not sufficient here: Classification predicts image-level categories while detection provides positions.

Question 344

The approved design action is: Preserve table row-column structure and validate numeric totals. Which test result would BEST establish that it worked?

  1. The answer resolves to a specific approved source document version
  2. Retrieved chunks correspond to coherent headings and source pages
  3. Extracted line items reconcile to visible row amounts
  4. Total mismatches trigger a review rather than automatic posting

Explanation

Correct Answer

C:

Cell relationships are necessary to check item totals. The expected acceptance evidence is: Extracted line items reconcile to visible row amounts. This observation directly tests the stated behavior.

Incorrect Answers

A:

Verifiable provenance differentiates a cited answer from guesswork. This observation does not establish the required result: Extracted line items reconcile to visible row amounts.

B:

RAG preparation should preserve document structure and source references. This observation does not establish the required result: Extracted line items reconcile to visible row amounts.

D:

Independent business rules catch plausible extraction errors. This observation does not establish the required result: Extracted line items reconcile to visible row amounts.

Question 345

Troubleshooting report: The client sends a filename as plain text to a text-only model. Which correction is MOST directly targeted to this failure?

  1. Create and test a simple one-agent tool workflow
  2. Prune irrelevant history and preserve the minimum required state
  3. Provision proper identity and resource role bindings for the target tenant
  4. Use the supported image-content input and multimodal deployment

Explanation

Correct Answer

D:

This correction directly addresses the reported failure. The model must receive the actual visual content to reason about it. The targeted change is: Use the supported image-content input and multimodal deployment.

Incorrect Answers

A:

A single agent can orchestrate a bounded tool without a multi-agent system. In the stated case, troubleshooting report: The client sends a filename as plain text to a text-only model; “Create and test a simple one-agent tool workflow” does not target that failure.

B:

Long irrelevant context consumes tokens and may increase latency. In the stated case, troubleshooting report: The client sends a filename as plain text to a text-only model; “Prune irrelevant history and preserve the minimum required state” does not target that failure.

C:

Copying code does not grant access to another tenant's resources. In the stated case, troubleshooting report: The client sends a filename as plain text to a text-only model; “Provision proper identity and resource role bindings for the target tenant” does not target that failure.

Question 346

Troubleshooting report: Clients immediately retry all throttled requests in parallel. Which correction is MOST directly targeted to this failure?

  1. Create a supported deployment and confirm resource access
  2. Rewrite the prompt with explicit acceptance criteria
  3. Exercise negative and expected tool paths before release
  4. Respect retry guidance and cap concurrent calls

Explanation

Correct Answer

D:

This correction directly addresses the reported failure. Rate limits protect the service and uncontrolled retries amplify overload. The targeted change is: Respect retry guidance and cap concurrent calls.

Incorrect Answers

A:

The portal provides model selection, deployment and test interfaces for supported resources. In the stated case, troubleshooting report: Clients immediately retry all throttled requests in parallel; “Create a supported deployment and confirm resource access” does not target that failure.

B:

Specific instructions reduce ambiguity about what constitutes success. In the stated case, troubleshooting report: Clients immediately retry all throttled requests in parallel; “Rewrite the prompt with explicit acceptance criteria” does not target that failure.

C:

An answer that sounds plausible may conceal failed tool calls. In the stated case, troubleshooting report: Clients immediately retry all throttled requests in parallel; “Exercise negative and expected tool paths before release” does not target that failure.

Question 347

An acceptance tester needs to validate: Enforce data classification and authorization before tool invocation. Which observed result proves the intended behavior?

  1. A simulated timeout yields pending status with a safe retry or escalation
  2. Sensitive fields are blocked from unapproved destinations in negative tests
  3. Browser traffic cannot access the protected model credential
  4. The target model deploys in the approved region and meets residency controls

Explanation

Correct Answer

B:

Tool descriptions cannot override data-loss-prevention boundaries. The expected acceptance evidence is: Sensitive fields are blocked from unapproved destinations in negative tests. This observation directly tests the stated behavior.

Incorrect Answers

A:

Timeout handling should not be reported as a successful action. This observation does not establish the required result: Sensitive fields are blocked from unapproved destinations in negative tests.

C:

Backend controls protect credentials, quotas and data policy. This observation does not establish the required result: Sensitive fields are blocked from unapproved destinations in negative tests.

D:

Not every model or feature is available in every location. This observation does not establish the required result: Sensitive fields are blocked from unapproved destinations in negative tests.

Question 348

An Azure-based AI team must satisfy this objective: A customer support recording has multiple speakers and needs attributed words. Which design choice is best?

  1. Use speech-to-text recognition on a supported audio input stream
  2. Use Speech Synthesis Markup Language pauses or breaks where supported
  3. Use a supported diarization or speaker-labeling capability where available
  4. Stop active recording and apply the documented microphone permission state

Explanation

Correct Answer

C:

Speaker diarization separates turns and identities at a conversation level. This implementation addresses the precise goal rather than an adjacent AI task.

Incorrect Answers

A:

Speech recognition converts waveform features to text tokens. That is not sufficient here: Speaker diarization separates turns and identities at a conversation level.

B:

SSML controls aspects of prosody and timing for synthesized speech. That is not sufficient here: Speaker diarization separates turns and identities at a conversation level.

D:

User control of capture must be enforced at the device or application layer. That is not sufficient here: Speaker diarization separates turns and identities at a conversation level.

Question 349

Troubleshooting report: The service returns only an overall positive attitude. Which correction is MOST directly targeted to this failure?

  1. Add the suitable pronunciation markup or lexicon
  2. Supply an authorized identity and matching resource configuration
  3. Keep operation outputs separate and map each to its intended field
  4. Use entity extraction instead of sentiment classification

Explanation

Correct Answer

D:

This correction directly addresses the reported failure. NER identifies named spans rather than summarizing the email. The targeted change is: Use entity extraction instead of sentiment classification.

Incorrect Answers

A:

SSML can guide pronunciation beyond plain-text guessing. In the stated case, troubleshooting report: The service returns only an overall positive attitude; “Add the suitable pronunciation markup or lexicon” does not target that failure.

B:

Audio capture success does not establish remote service authentication. In the stated case, troubleshooting report: The service returns only an overall positive attitude; “Supply an authorized identity and matching resource configuration” does not target that failure.

C:

Sentiment polarity and named entity spans answer different questions. In the stated case, troubleshooting report: The service returns only an overall positive attitude; “Keep operation outputs separate and map each to its intended field” does not target that failure.

Question 350

An acceptance tester needs to validate: Configure the client with the actual resource endpoint and deployment identifier. Which observed result proves the intended behavior?

  1. A load test recovers without a retry storm or duplicate effects
  2. Conflict tests keep disallowed actions blocked
  3. A request using the deployment name returns a response from that resource
  4. The response contains only permitted fields with correct types

Explanation

Correct Answer

C:

An inferencing request is routed to a concrete deployment configured in the resource. The expected acceptance evidence is: A request using the deployment name returns a response from that resource. This observation directly tests the stated behavior.

Incorrect Answers

A:

Rate limits protect the service and uncontrolled retries amplify overload. This observation does not establish the required result: A request using the deployment name returns a response from that resource.

B:

Developer and system constraints govern allowed operations over user requests. This observation does not establish the required result: A request using the deployment name returns a response from that resource.

D:

Model output constraints help downstream parsing but still require validation. This observation does not establish the required result: A request using the deployment name returns a response from that resource.

Microsoft Azure AI Fundamentals (AI-901) Practice Tests