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 68

Question 670

A project needs the following AI behavior: A team wants a system prompt that constrains a support assistant to approved policies. What should the designer implement?

  1. Store the approved instruction as versioned application configuration
  2. Put persistent task boundaries in a supported high-priority system instruction
  3. State the role, goal, provided context, output requirement and constraints clearly
  4. Create one agent in the Foundry portal with scoped tool definitions

Explanation

Correct Answer

B:

High-priority instructions should describe approved behavior and escalation rules. This implementation addresses the precise goal rather than an adjacent AI task.

Incorrect Answers

A:

Versioned system messages provide a controlled behavioral baseline. That is not sufficient here: High-priority instructions should describe approved behavior and escalation rules.

C:

Specific instructions reduce ambiguity about what constitutes success. That is not sufficient here: High-priority instructions should describe approved behavior and escalation rules.

D:

A single agent can orchestrate a bounded tool without a multi-agent system. That is not sufficient here: High-priority instructions should describe approved behavior and escalation rules.

Question 671

An engineering review documents this requirement: A training voice assistant needs correct pronunciation of a product acronym. Select the most appropriate solution.

  1. Use supported SSML phoneme or say-as controls for pronunciation
  2. Use speech-to-text recognition on a supported audio input stream
  3. Check synthesized voice availability for the selected locale and region
  4. Configure a supported Spanish recognition locale and test representative speakers

Explanation

Correct Answer

A:

SSML can guide pronunciation beyond plain-text guessing. This implementation addresses the precise goal rather than an adjacent AI task.

Incorrect Answers

B:

Speech recognition converts waveform features to text tokens. That is not sufficient here: SSML can guide pronunciation beyond plain-text guessing.

C:

Not every voice is supported for all languages. That is not sufficient here: SSML can guide pronunciation beyond plain-text guessing.

D:

Recognition models need the appropriate supported language settings. That is not sufficient here: SSML can guide pronunciation beyond plain-text guessing.

Question 672

An Azure AI Fundamentals solution has this requirement: A user asks what a photo shows overall. Which implementation most directly addresses it?

  1. Apply supported image safety filtering and review before distribution
  2. Select appropriate separate visual features for OCR, tagging and people detection
  3. Request an image caption or supported multimodal description
  4. Preserve region coordinates with extracted text or field evidence

Explanation

Correct Answer

C:

Captioning produces a sentence-level description rather than a coordinate list. This implementation addresses the precise goal rather than an adjacent AI task.

Incorrect Answers

A:

Generative media needs policy controls in addition to prompt instructions. That is not sufficient here: Captioning produces a sentence-level description rather than a coordinate list.

B:

Different visual outputs describe text, scene classes and detected people. That is not sufficient here: Captioning produces a sentence-level description rather than a coordinate list.

D:

Spatial provenance supports later verification of recognition results. That is not sufficient here: Captioning produces a sentence-level description rather than a coordinate list.

Question 673

Troubleshooting report: The app keeps using a service principal authorized only in the first tenant. Which correction is MOST directly targeted to this failure?

  1. Provision proper identity and resource role bindings for the target tenant
  2. Restrict the tool to approved documents and identities
  3. Add approval state and verified role checks
  4. Tighten the output contract and apply post-response validation

Explanation

Correct Answer

A:

This correction directly addresses the reported failure. Copying code does not grant access to another tenant's resources. The targeted change is: Provision proper identity and resource role bindings for the target tenant.

Incorrect Answers

B:

Tool descriptions guide selection, while backend rules enforce allowed sources. In the stated case, troubleshooting report: The app keeps using a service principal authorized only in the first tenant; “Restrict the tool to approved documents and identities” does not target that failure.

C:

High-impact side effects need permissions independent of conversational language. In the stated case, troubleshooting report: The app keeps using a service principal authorized only in the first tenant; “Add approval state and verified role checks” does not target that failure.

D:

Explicit examples can improve adherence to bounded output conventions. In the stated case, troubleshooting report: The app keeps using a service principal authorized only in the first tenant; “Tighten the output contract and apply post-response validation” does not target that failure.

Question 674

Troubleshooting report: One thumbnail is used as the basis for the entire shift summary. Which correction is MOST directly targeted to this failure?

  1. Analyze multiple temporal segments and preserve timecodes
  2. Use translation and inspect named terms
  3. Invoke image segmentation and compositing or a controlled image editor
  4. Treat absent fields as unknown and request a better image

Explanation

Correct Answer

A:

This correction directly addresses the reported failure. Video understanding retains order and time, unlike single-image tagging. The targeted change is: Analyze multiple temporal segments and preserve timecodes.

Incorrect Answers

B:

Translation transfers meaning across languages rather than changing topic labels. In the stated case, troubleshooting report: One thumbnail is used as the basis for the entire shift summary; “Use translation and inspect named terms” does not target that failure.

C:

Editing changes selected pixels while leaving product regions intact. In the stated case, troubleshooting report: One thumbnail is used as the basis for the entire shift summary; “Invoke image segmentation and compositing or a controlled image editor” does not target that failure.

D:

Images of receipts need OCR/layout extraction before typed business values. In the stated case, troubleshooting report: One thumbnail is used as the basis for the entire shift summary; “Treat absent fields as unknown and request a better image” does not target that failure.

Question 675

An Azure AI solution uses the following approved control: Use supported speaker/timing evidence and keep uncertain attribution explicit. What observable result should pass QA?

  1. A reply links its claim to a transcript time or relevant video segment
  2. Observed actions link to supported source intervals while unobserved short events are not claimed as detected
  3. Speaker assignments align to audible turns with flagged overlaps
  4. Extracted rows align with the original form cells

Explanation

Correct Answer

C:

Conversation extraction needs more than a single flattened text stream. The expected acceptance evidence is: Speaker assignments align to audible turns with flagged overlaps. This observation directly tests the stated behavior.

Incorrect Answers

A:

Multimodal grounding should preserve how the answer maps to source media. This observation does not establish the required result: Speaker assignments align to audible turns with flagged overlaps.

B:

Content Understanding samples approximately one frame per second, so brief or single-frame actions can be missed despite valid timestamps. This observation does not establish the required result: Speaker assignments align to audible turns with flagged overlaps.

D:

Ordinary plain-text OCR may lose table relationships and checkbox states. This observation does not establish the required result: Speaker assignments align to audible turns with flagged overlaps.

Question 676

An Azure-based AI team must satisfy this objective: A public service assistant is used by people with screen readers. Which design choice is best?

  1. Provide semantic labels and keyboard-operable controls for all important interactions
  2. Test accuracy on representative lighting and camera conditions and provide a safe fallback
  3. Set explicit uncertainty thresholds and route high-risk decisions to qualified review
  4. Require governance evidence with reviewers, test outcomes, and documented exceptions

Explanation

Correct Answer

A:

Inclusive design permits nonvisual navigation of the same task path. This implementation addresses the precise goal rather than an adjacent AI task.

Incorrect Answers

B:

Reliability demands testing the actual operating environment, not just clean samples. That is not sufficient here: Inclusive design permits nonvisual navigation of the same task path.

C:

Safety-critical recommendations require verified constraints outside open-ended generation. That is not sufficient here: Inclusive design permits nonvisual navigation of the same task path.

D:

Self-attestation is not independent proof of regulatory controls. That is not sufficient here: Inclusive design permits nonvisual navigation of the same task path.

Question 677

Troubleshooting report: The team removes one sensitive input and assumes bias is gone. Which correction is MOST directly targeted to this failure?

  1. Review label provenance and downstream decision disparities
  2. Re-embed the corpus or use a versioned compatible index
  3. Refresh indexed grounding sources and verify their effective date
  4. Select a classification output for image-wide categories

Explanation

Correct Answer

A:

This correction directly addresses the reported failure. Bias can enter through the targets themselves, not just inputs. The targeted change is: Review label provenance and downstream decision disparities.

Incorrect Answers

B:

Incompatible vector spaces can make nearest-neighbor rankings meaningless. In the stated case, troubleshooting report: The team removes one sensitive input and assumes bias is gone; “Re-embed the corpus or use a versioned compatible index” does not target that failure.

C:

External retrieval updates factual context without retraining model weights. In the stated case, troubleshooting report: The team removes one sensitive input and assumes bias is gone; “Refresh indexed grounding sources and verify their effective date” does not target that failure.

D:

Classification assigns image-wide categories rather than bounding-box coordinates. In the stated case, troubleshooting report: The team removes one sensitive input and assumes bias is gone; “Select a classification output for image-wide categories” does not target that failure.

Question 678

An engineering review documents this requirement: A vision model fails in low-light conditions encountered by warehouse night staff. Select the most appropriate solution.

  1. Test representative document layouts and accessibility-oriented inputs
  2. Ask for clarification or abstain rather than acting on inferred intent
  3. Redact personal fields before telemetry storage and control log access
  4. Test accuracy on representative lighting and camera conditions and provide a safe fallback

Explanation

Correct Answer

D:

Reliability demands testing the actual operating environment, not just clean samples. This implementation addresses the precise goal rather than an adjacent AI task.

Incorrect Answers

A:

Inclusion requires robust input handling beyond a single document template. That is not sufficient here: Reliability demands testing the actual operating environment, not just clean samples.

B:

Reliable assistance distinguishes missing intent from authorized action. That is not sufficient here: Reliability demands testing the actual operating environment, not just clean samples.

C:

Troubleshooting does not justify reproducing sensitive application records. That is not sufficient here: Reliability demands testing the actual operating environment, not just clean samples.

Question 679

A solution team has selected this control: Minimize extracted PII and restrict diagnostic payload storage. Which observation is the most direct acceptance evidence?

  1. The output identifies both amount and confirmed currency
  2. Authorized structured fields remain but raw IDs do not appear in logs
  3. Each known document type reaches the expected analyzer
  4. A question links to the correct contract section and page

Explanation

Correct Answer

B:

Inference outputs and telemetry can contain protected identity information. The expected acceptance evidence is: Authorized structured fields remain but raw IDs do not appear in logs. This observation directly tests the stated behavior.

Incorrect Answers

A:

A currency symbol or amount alone may not establish financial meaning. This observation does not establish the required result: Authorized structured fields remain but raw IDs do not appear in logs.

C:

Different document kinds have distinct field schemas and review paths. This observation does not establish the required result: Authorized structured fields remain but raw IDs do not appear in logs.

D:

Structured layout and provenance let users verify a quoted clause. This observation does not establish the required result: Authorized structured fields remain but raw IDs do not appear in logs.

Question 680

A project needs the following AI behavior: An AI agent reports that a refund succeeded even though the payment service returned an error. What should the designer implement?

  1. Test representative document layouts and accessibility-oriented inputs
  2. Assign accountable operational and risk owners with escalation responsibilities
  3. Verify an authoritative transaction receipt before claiming task completion
  4. Provide synchronized transcripts or captions for spoken content

Explanation

Correct Answer

C:

Generated text is not evidence of external side effects. This implementation addresses the precise goal rather than an adjacent AI task.

Incorrect Answers

A:

Inclusion requires robust input handling beyond a single document template. That is not sufficient here: Generated text is not evidence of external side effects.

B:

An AI model cannot be its own accountable incident responder. That is not sufficient here: Generated text is not evidence of external side effects.

D:

Accessible information must not depend on hearing audio. That is not sufficient here: Generated text is not evidence of external side effects.

Microsoft Azure AI Fundamentals (AI-901) Practice Tests