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 82

Question 810

An AI operations team is testing two distinct capabilities. Failure (1): The summarizer uses only a single thumbnail to infer all events. Failure (2): The search index exposes all extracted pages to every client. Which TWO specific corrections should the team apply?

  1. Enforce source permissions in indexing and query-time filtering
  2. Represent repeated fields as arrays and validate counts
  3. Inspect multiple temporal segments and independently review brief events that sampled frames could miss
  4. Build citations from stored IDs and source spans
  5. Keep transcript anchors and validate extracted tasks

Explanation

Correct Answers

A:

Correct for failure (2). Extracting content does not waive document access controls. It independently corrects the second defect.

C:

Correct for failure (1). Content Understanding samples approximately one frame per second, so brief or single-frame actions can be missed despite valid timestamps. The proposed correction directly addresses that reported defect.

Incorrect Answers

B:

Repeated records require structured lists with item-level associations. It does not complete the two decisions required in this question.

D:

Verifiable provenance differentiates a cited answer from guesswork. It does not complete the two decisions required in this question.

E:

Transcription and summarization are separate results with different validation. The stated pair of failures instead requires “Enforce source permissions in indexing and query-time filtering” and “Inspect multiple temporal segments and independently review brief events that sampled frames could miss”; this option cannot substitute for both.

Question 811

An engineering review documents this requirement: An agent retries a non-idempotent order action after a network timeout. Select the most appropriate solution.

  1. Use operation IDs and read-back receipts before replaying external side effects
  2. Externalize endpoint and deployment configuration per environment
  3. Check regional model availability and data handling requirements before deployment
  4. Use typed state fields and validate identifiers at each tool boundary

Explanation

Correct Answer

A:

A timeout does not prove that the original mutation failed. This implementation addresses the precise goal rather than an adjacent AI task.

Incorrect Answers

B:

A correct code path can still target the wrong resource after migration. That is not sufficient here: A timeout does not prove that the original mutation failed.

C:

Not every model or feature is available in every location. That is not sufficient here: A timeout does not prove that the original mutation failed.

D:

Structured state prevents subtle substitution of unrelated customer records. That is not sufficient here: A timeout does not prove that the original mutation failed.

Question 812

A project needs the following AI behavior: An application sends simultaneous Foundry requests for different customers. What should the designer implement?

  1. Build a supported messages collection with roles and the user's content
  2. Use operation IDs and read-back receipts before replaying external side effects
  3. Use managed identity or protected secret storage rather than embedding keys in code
  4. Assign each request its own authenticated context and correlate responses to request IDs

Explanation

Correct Answer

D:

Concurrent inference must not mix user data or response ownership. This implementation addresses the precise goal rather than an adjacent AI task.

Incorrect Answers

A:

Chat APIs use structured messages rather than concatenating untyped command strings. That is not sufficient here: Concurrent inference must not mix user data or response ownership.

B:

A timeout does not prove that the original mutation failed. That is not sufficient here: Concurrent inference must not mix user data or response ownership.

C:

Credentials should not appear in repositories, prompt logs or client binaries. That is not sufficient here: Concurrent inference must not mix user data or response ownership.

Question 813

An acceptance tester needs to validate: Use retrieval-augmented generation from current approved catalog sources. Which observed result proves the intended behavior?

  1. Repeated structured cases follow the expected schema more consistently
  2. A new approved price appears in answers after source refresh
  3. A regression test records model version and key generation settings
  4. Known arithmetic cases agree exactly with verified calculations

Explanation

Correct Answer

B:

External retrieval updates factual context without retraining model weights. The expected acceptance evidence is: A new approved price appears in answers after source refresh. This observation directly tests the stated behavior.

Incorrect Answers

A:

Lower randomness often improves consistency but does not guarantee correctness. This observation does not establish the required result: A new approved price appears in answers after source refresh.

C:

Reproducibility depends on more than the visible prompt text. This observation does not establish the required result: A new approved price appears in answers after source refresh.

D:

Generative models are not a substitute for auditable deterministic computation. This observation does not establish the required result: A new approved price appears in answers after source refresh.

Question 814

Troubleshooting report: An image classifier only returns the category vehicle. Which correction is MOST directly targeted to this failure?

  1. Use a suitable video analyzer with supported source format
  2. Configure required invoice fields and validate extracted values
  3. Run a supported OCR or visual text extraction feature
  4. Preserve conflict flags and apply approved precedence rules

Explanation

Correct Answer

C:

This correction directly addresses the reported failure. OCR is appropriate for characters embedded in a photograph. The targeted change is: Run a supported OCR or visual text extraction feature.

Incorrect Answers

A:

Multimodal extraction draws on visual and spoken evidence over time. In the stated case, troubleshooting report: An image classifier only returns the category vehicle; “Use a suitable video analyzer with supported source format” does not target that failure.

B:

Document extraction turns unstructured pages into validated structured fields. In the stated case, troubleshooting report: An image classifier only returns the category vehicle; “Configure required invoice fields and validate extracted values” does not target that failure.

D:

A model should not silently choose among competing contractual evidence. In the stated case, troubleshooting report: An image classifier only returns the category vehicle; “Preserve conflict flags and apply approved precedence rules” does not target that failure.

Question 815

The approved design action is: Adjust supported sampling parameters while preserving safety and quality tests. Which test result would BEST establish that it worked?

  1. A held-out set receives the correct whole-image classes and empirically calibrated class probabilities
  2. Reintroducing relevant context changes the answer in a controlled test
  3. The output returns evaluated sentiment classes rather than generated images
  4. The same brief yields a wider range of acceptable styles without broken constraints

Explanation

Correct Answer

D:

More varied token sampling can increase creative alternatives. The expected acceptance evidence is: The same brief yields a wider range of acceptable styles without broken constraints. This observation directly tests the stated behavior.

Incorrect Answers

A:

Classification assigns image-wide categories rather than bounding-box coordinates. This observation does not establish the required result: The same brief yields a wider range of acceptable styles without broken constraints.

B:

A model conditions its output on supplied instructions and conversation text. This observation does not establish the required result: The same brief yields a wider range of acceptable styles without broken constraints.

C:

Different model tasks require compatible outputs and labeled evaluation data. This observation does not establish the required result: The same brief yields a wider range of acceptable styles without broken constraints.

Question 816

A project needs the following AI behavior: A team must use an audio-capable model for speech understanding but receives a text-only response. What should the designer implement?

  1. Evaluate word error rate on the same held-out audio samples
  2. Check the deployment supports the requested audio-input modality and request structure
  3. Send the audio through a deployment that supports audio inputs
  4. Use supported voice and SSML prosody rate controls

Explanation

Correct Answer

B:

Model capabilities and accepted payload types differ across deployments. This implementation addresses the precise goal rather than an adjacent AI task.

Incorrect Answers

A:

Comparable labeled audio prevents misleading quality claims. That is not sufficient here: Model capabilities and accepted payload types differ across deployments.

C:

A text-only request cannot interpret sound content by file name alone. That is not sufficient here: Model capabilities and accepted payload types differ across deployments.

D:

Voice selection and prosody can influence delivery without changing the text. That is not sufficient here: Model capabilities and accepted payload types differ across deployments.

Question 817

An engineering review documents this requirement: A team wants answers matching a particular corporate writing style. Select the most appropriate solution.

  1. Provide tested style instructions and examples in the prompt before considering tuning
  2. Use clustering or another appropriate unsupervised technique
  3. Use a held-out evaluation and investigate overfitting or data leakage
  4. Require verifiable grounding and calibrated uncertainty

Explanation

Correct Answer

A:

Clear prompting can control tone without model retraining. This implementation addresses the precise goal rather than an adjacent AI task.

Incorrect Answers

B:

Unsupervised learning finds structure without supplied target labels. That is not sufficient here: Clear prompting can control tone without model retraining.

C:

A model can memorize training examples without generalizing. That is not sufficient here: Clear prompting can control tone without model retraining.

D:

Language-model likelihood does not independently establish factual truth. That is not sufficient here: Clear prompting can control tone without model retraining.

Question 818

An AI operations team is testing two distinct capabilities. Failure (1): The system only translates a typed transcript after the conversation ends. Failure (2): The model labels the entire image as a store without locating items. Which TWO specific corrections should the team apply?

  1. Use a detection output with bounding boxes
  2. Require source-linked summarization evaluation
  3. Replace the TTS stage with recognition/transcription
  4. Analyze multiple temporal segments and preserve timecodes
  5. Use a near-real-time speech translation pipeline

Explanation

Correct Answers

A:

Correct for failure (2). Object detectors return classes and spatial boundaries. It independently corrects the second defect.

E:

Correct for failure (1). Speech translation spans audio input, semantic transfer and optional output audio. The proposed correction directly addresses that reported defect.

Incorrect Answers

B:

Summarization compresses content without independently validating missing facts. It does not complete the two decisions required in this question.

C:

Speech recognition converts speech into words suitable for search. The stated pair of failures instead requires “Use a detection output with bounding boxes” and “Use a near-real-time speech translation pipeline”; this option cannot substitute for both.

D:

Video understanding retains order and time, unlike single-image tagging. The stated pair of failures instead requires “Use a detection output with bounding boxes” and “Use a near-real-time speech translation pipeline”; this option cannot substitute for both.

Question 819

Troubleshooting report: The code appends image tags to extracted invoice number fields. Which correction is MOST directly targeted to this failure?

  1. Move sensitive images to authenticated private storage
  2. Request the needed visual features explicitly
  3. Parse outputs by their documented feature-specific schema
  4. Constrain the prompt and validate the final artwork

Explanation

Correct Answer

C:

This correction directly addresses the reported failure. Tags describe concepts while OCR reports character content. The targeted change is: Parse outputs by their documented feature-specific schema.

Incorrect Answers

A:

Images can carry personally identifying information beyond text fields. In the stated case, troubleshooting report: The code appends image tags to extracted invoice number fields; “Move sensitive images to authenticated private storage” does not target that failure.

B:

Different visual outputs describe text, scene classes and detected people. In the stated case, troubleshooting report: The code appends image tags to extracted invoice number fields; “Request the needed visual features explicitly” does not target that failure.

D:

Explicit visual guidance can influence generated appearance but is not absolute. In the stated case, troubleshooting report: The code appends image tags to extracted invoice number fields; “Constrain the prompt and validate the final artwork” does not target that failure.

Question 820

Troubleshooting report: The output accidentally includes content from an unrelated uploaded image. Which correction is MOST directly targeted to this failure?

  1. Use image input and OCR or vision grounding
  2. Route to the text-to-image mode with intended inputs only
  3. Request the needed visual features explicitly
  4. Gate distribution on the image moderation result

Explanation

Correct Answer

B:

This correction directly addresses the reported failure. Generation from text is distinct from image editing. The targeted change is: Route to the text-to-image mode with intended inputs only.

Incorrect Answers

A:

Visual evidence is needed to read sign content. In the stated case, troubleshooting report: The output accidentally includes content from an unrelated uploaded image; “Use image input and OCR or vision grounding” does not target that failure.

C:

Different visual outputs describe text, scene classes and detected people. In the stated case, troubleshooting report: The output accidentally includes content from an unrelated uploaded image; “Request the needed visual features explicitly” does not target that failure.

D:

Generative media needs policy controls in addition to prompt instructions. In the stated case, troubleshooting report: The output accidentally includes content from an unrelated uploaded image; “Gate distribution on the image moderation result” does not target that failure.

Microsoft Azure AI Fundamentals (AI-901) Practice Tests