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 70

Question 690

An Azure-based AI team must satisfy this objective: A shop wants to automatically replace a green product background with white. Which design choice is best?

  1. Use machine translation with locale-specific quality review
  2. Use clustering with normalized behavioral features
  3. Use image editing or segmentation with object-mask preservation
  4. Use text classification with labeled intent examples

Explanation

Correct Answer

C:

Editing changes selected pixels while leaving product regions intact. This implementation addresses the precise goal rather than an adjacent AI task.

Incorrect Answers

A:

Translation transfers meaning across languages rather than changing topic labels. That is not sufficient here: Editing changes selected pixels while leaving product regions intact.

B:

Clustering discovers groups based on feature similarity. That is not sufficient here: Editing changes selected pixels while leaving product regions intact.

D:

Classification maps text to one of several predefined categories. That is not sufficient here: Editing changes selected pixels while leaving product regions intact.

Question 691

The approved design action is: Evaluate safety again after tool and retrieval results are incorporated. Which test result would BEST establish that it worked?

  1. Eligible memory entries disappear from active stores after the verified request
  2. An unprivileged test user is denied protected operations despite claiming to be an admin
  3. Out-of-scope questions trigger qualified language and reference suggestions
  4. A prompt-injection test remains unable to trigger a prohibited tool after retrieval

Explanation

Correct Answer

D:

Threats can enter later stages even when the original user request was harmless. The expected acceptance evidence is: A prompt-injection test remains unable to trigger a prohibited tool after retrieval. This observation directly tests the stated behavior.

Incorrect Answers

A:

Accountability requires evidence of actual data lifecycle actions. This observation does not establish the required result: A prompt-injection test remains unable to trigger a prohibited tool after retrieval.

B:

Natural-language claims cannot elevate authentication authority. This observation does not establish the required result: A prompt-injection test remains unable to trigger a prohibited tool after retrieval.

C:

Honesty requires acknowledging when a claim cannot be substantiated. This observation does not establish the required result: A prompt-injection test remains unable to trigger a prohibited tool after retrieval.

Question 692

A solution team has selected this control: Use a relevance-ranking or recommendation model evaluated on resolved-case history. Which observation is the most direct acceptance evidence?

  1. The spoken message matches the approved written text
  2. The summary preserves the topic sequence and supported key points
  3. Held-out events produce measured recall and precision for failures
  4. The top ranked article matches the issue and is permitted for that customer

Explanation

Correct Answer

D:

Recommendations rank candidate items rather than simply generating arbitrary titles. The expected acceptance evidence is: The top ranked article matches the issue and is permitted for that customer. This observation directly tests the stated behavior.

Incorrect Answers

A:

Speech synthesis converts supplied text into audible output. This observation does not establish the required result: The top ranked article matches the issue and is permitted for that customer.

B:

Meeting summaries require temporal and spoken source content. This observation does not establish the required result: The top ranked article matches the issue and is permitted for that customer.

C:

A failure probability or class is a discrete decision under a defined horizon. This observation does not establish the required result: The top ranked article matches the issue and is permitted for that customer.

Question 693

An Azure AI solution uses the following approved control: Use an appropriate Content Understanding or invoice-specific analyzer with a typed field schema. What observable result should pass QA?

  1. A test invoice yields correct merchant, date and total with source anchors
  2. A reply links its claim to a transcript time or relevant video segment
  3. A restricted user cannot search another tenant's contract
  4. Ambiguous totals are flagged for manual review with source crops

Explanation

Correct Answer

A:

Document extraction turns unstructured pages into validated structured fields. The expected acceptance evidence is: A test invoice yields correct merchant, date and total with source anchors. This observation directly tests the stated behavior.

Incorrect Answers

B:

Multimodal grounding should preserve how the answer maps to source media. This observation does not establish the required result: A test invoice yields correct merchant, date and total with source anchors.

C:

Extracting content does not waive document access controls. This observation does not establish the required result: A test invoice yields correct merchant, date and total with source anchors.

D:

A low-confidence amount must not be silently treated as definitive. This observation does not establish the required result: A test invoice yields correct merchant, date and total with source anchors.

Question 694

An Azure AI Fundamentals solution has this requirement: A production deployment produces different output than a tested sandbox. Which implementation most directly addresses it?

  1. Compare model deployment, prompt version, retrieval index and tool configuration
  2. Limit retrieval and chat context to the authenticated case scope
  3. Give each tool a discriminating description and narrow permissions
  4. Use a versioned evaluation set with stable scoring and test inputs

Explanation

Correct Answer

A:

Environment drift can change behavior without code edits. This implementation addresses the precise goal rather than an adjacent AI task.

Incorrect Answers

B:

Source scoping enforces privacy and also improves answer relevance. That is not sufficient here: Environment drift can change behavior without code edits.

C:

Clear tool contracts reduce confusion between read and write operations. That is not sufficient here: Environment drift can change behavior without code edits.

D:

Anecdotal success on different prompts cannot establish an improvement. That is not sufficient here: Environment drift can change behavior without code edits.

Question 695

Troubleshooting report: The application returns only numeric object counts. Which correction is MOST directly targeted to this failure?

  1. Use image captioning with verified visual input
  2. Use aspect-ratio-preserving crop or padding
  3. Limit concurrency and honor retry guidance
  4. Add confidence gating and supporting evidence

Explanation

Correct Answer

A:

This correction directly addresses the reported failure. Captioning produces a sentence-level description rather than a coordinate list. The targeted change is: Use image captioning with verified visual input.

Incorrect Answers

B:

Smart cropping differs from simply stretching pixels. In the stated case, troubleshooting report: The application returns only numeric object counts; “Use aspect-ratio-preserving crop or padding” does not target that failure.

C:

Too many concurrent requests can exceed resource quota. In the stated case, troubleshooting report: The application returns only numeric object counts; “Limit concurrency and honor retry guidance” does not target that failure.

D:

Generated visual interpretations can be wrong even when phrased fluently. In the stated case, troubleshooting report: The application returns only numeric object counts; “Add confidence gating and supporting evidence” does not target that failure.

Question 696

The approved design action is: Preserve region coordinates with extracted text or field evidence. Which test result would BEST establish that it worked?

  1. The answer reproduces the visible sign text accurately
  2. A reviewer can locate the source field in the image from its coordinates
  3. The result depicts the prompt without implying original photo preservation
  4. A multi-object test returns boxes when location is required

Explanation

Correct Answer

B:

Spatial provenance supports later verification of recognition results. The expected acceptance evidence is: A reviewer can locate the source field in the image from its coordinates. This observation directly tests the stated behavior.

Incorrect Answers

A:

Visual evidence is needed to read sign content. This observation does not establish the required result: A reviewer can locate the source field in the image from its coordinates.

C:

Generation from text is distinct from image editing. This observation does not establish the required result: A reviewer can locate the source field in the image from its coordinates.

D:

Classification predicts image-level categories while detection provides positions. This observation does not establish the required result: A reviewer can locate the source field in the image from its coordinates.

Question 697

An AI operations team is testing two distinct capabilities. Failure (1): The system claims all outputs are approved without checking them. Failure (2): The system returns one image-wide classification label. Which TWO specific corrections should the team apply?

  1. Use detection instead of whole-image classification
  2. Apply a visual acceptance review and keep the strongest matching variants
  3. Improve the product mask and restrict the editable region
  4. Apply smart-crop suggestions or salient-region detection
  5. Map the task to a whole-image classifier

Explanation

Correct Answers

A:

Correct for failure (2). Object detection reports which objects appear and where they occur. It independently corrects the second defect.

B:

Correct for failure (1). Image-generation variability can be useful but does not guarantee suitability. The proposed correction directly addresses that reported defect.

Incorrect Answers

C:

This action would address another problem: the edit changes the product logo while replacing the background. A mask defines which image area may change. It does not resolve either specified failure.

D:

Smart cropping attempts to retain salient regions within target dimensions. The stated pair of failures instead requires “Use detection instead of whole-image classification” and “Apply a visual acceptance review and keep the strongest matching variants”; this option cannot substitute for both.

E:

Image-level classification is distinct from object localization. It does not complete the two decisions required in this question.

Question 698

An Azure AI Fundamentals solution has this requirement: A prompt is too vague and the assistant answers a different task. Which implementation most directly addresses it?

  1. Use the supported run or response status and verify its terminal state
  2. Align the tool argument names and types with the published function schema
  3. Specify factual boundaries and check generated claims against approved evidence
  4. State the role, goal, provided context, output requirement and constraints clearly

Explanation

Correct Answer

D:

Specific instructions reduce ambiguity about what constitutes success. This implementation addresses the precise goal rather than an adjacent AI task.

Incorrect Answers

A:

Submission is not the same as successful completion. That is not sufficient here: Specific instructions reduce ambiguity about what constitutes success.

B:

Invalid tool parameters break calls even if the intended operation is correct. That is not sufficient here: Specific instructions reduce ambiguity about what constitutes success.

C:

Prompt instructions must be paired with evidence verification for consequential facts. That is not sufficient here: Specific instructions reduce ambiguity about what constitutes success.

Question 699

A project needs the following AI behavior: An invoice-processing workflow must reject extracted totals that disagree with line items. What should the designer implement?

  1. Extract currency code and monetary amount as separate validated fields
  2. Apply authorized retrieval filters to extracted chunks and cited source text
  3. Define schema fields for document date, due date and signature date
  4. Recompute arithmetic and compare totals with reported invoice fields

Explanation

Correct Answer

D:

Independent business rules catch plausible extraction errors. This implementation addresses the precise goal rather than an adjacent AI task.

Incorrect Answers

A:

A currency symbol or amount alone may not establish financial meaning. That is not sufficient here: Independent business rules catch plausible extraction errors.

B:

Extracting content does not waive document access controls. That is not sufficient here: Independent business rules catch plausible extraction errors.

C:

Typed extraction avoids mixing semantically different dates. That is not sufficient here: Independent business rules catch plausible extraction errors.

Question 700

Troubleshooting report: The prototype launches unnecessary specialist agents for one lookup. Which correction is MOST directly targeted to this failure?

  1. Move system instructions into protected application configuration
  2. Exercise negative and expected tool paths before release
  3. Apply case and role filters before providing model context
  4. Create and test a simple one-agent tool workflow

Explanation

Correct Answer

D:

This correction directly addresses the reported failure. A single agent can orchestrate a bounded tool without a multi-agent system. The targeted change is: Create and test a simple one-agent tool workflow.

Incorrect Answers

A:

Versioned system messages provide a controlled behavioral baseline. In the stated case, troubleshooting report: The prototype launches unnecessary specialist agents for one lookup; “Move system instructions into protected application configuration” does not target that failure.

B:

An answer that sounds plausible may conceal failed tool calls. In the stated case, troubleshooting report: The prototype launches unnecessary specialist agents for one lookup; “Exercise negative and expected tool paths before release” does not target that failure.

C:

Source scoping enforces privacy and also improves answer relevance. In the stated case, troubleshooting report: The prototype launches unnecessary specialist agents for one lookup; “Apply case and role filters before providing model context” does not target that failure.

Microsoft Azure AI Fundamentals (AI-901) Practice Tests