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 32

Question 310

Troubleshooting report: The agent executes an irreversible tool call from an incomplete command. Which correction is MOST directly targeted to this failure?

  1. Apply real sign-in and role checks before invoking protected tools
  2. Run the frozen adversarial and ordinary acceptance suite against the candidate
  3. Enforce backend authorization and signed approval regardless of prompt wording
  4. Put an explicit confirmation gate before high-impact actions

Explanation

Correct Answer

D:

This correction directly addresses the reported failure. Reliable assistance distinguishes missing intent from authorized action. The targeted change is: Put an explicit confirmation gate before high-impact actions.

Incorrect Answers

A:

Natural-language claims cannot elevate authentication authority. In the stated case, troubleshooting report: The agent executes an irreversible tool call from an incomplete command; “Apply real sign-in and role checks before invoking protected tools” does not target that failure.

B:

Reliability includes preventing behavior regression across model versions. In the stated case, troubleshooting report: The agent executes an irreversible tool call from an incomplete command; “Run the frozen adversarial and ordinary acceptance suite against the candidate” does not target that failure.

C:

Accountable high-impact operations need verifiable authorized signatures. In the stated case, troubleshooting report: The agent executes an irreversible tool call from an incomplete command; “Enforce backend authorization and signed approval regardless of prompt wording” does not target that failure.

Question 311

An Azure-based AI team must satisfy this objective: An HR department must route incoming PDFs by document type before extraction. Which design choice is best?

  1. Use noise-aware speech recognition and intent classification with fallback
  2. Use invoice-oriented structured extraction with table row preservation
  3. Use image interpretation followed by grounded text and speech synthesis
  4. Use document classification followed by the appropriate analyzer

Explanation

Correct Answer

D:

Classification selects the extraction process matching the document category. This implementation addresses the precise goal rather than an adjacent AI task.

Incorrect Answers

A:

Speech input quality affects subsequent language classification. That is not sufficient here: Classification selects the extraction process matching the document category.

B:

Table extraction associates quantities, descriptions and prices by row. That is not sufficient here: Classification selects the extraction process matching the document category.

C:

Multimodal input can describe the graphic and speech output can read that description. That is not sufficient here: Classification selects the extraction process matching the document category.

Question 312

An Azure-based AI team must satisfy this objective: An intake processor receives forms with multiple dates. Which design choice is best?

  1. Define schema fields for document date, due date and signature date
  2. Classify document type before routing to the relevant extraction analyzer
  3. Preserve document IDs, versions and extracted field evidence with search results
  4. Create a schema for policy type, coverage amount and effective dates

Explanation

Correct Answer

A:

Typed extraction avoids mixing semantically different dates. This implementation addresses the precise goal rather than an adjacent AI task.

Incorrect Answers

B:

Different document kinds have distinct field schemas and review paths. That is not sufficient here: Typed extraction avoids mixing semantically different dates.

C:

Verifiable provenance differentiates a cited answer from guesswork. That is not sufficient here: Typed extraction avoids mixing semantically different dates.

D:

Custom extraction maps semantically relevant phrases to typed fields. That is not sufficient here: Typed extraction avoids mixing semantically different dates.

Question 313

A solution team has selected this control: Use OCR tuned to the source quality and validate extracted text. Which observation is the most direct acceptance evidence?

  1. A test correctly references the specific on-screen control
  2. Extracted checks and notes match visible form locations
  3. A prompt-only request yields a new original illustration rather than a classification label
  4. Handwritten serials are returned with confidence for review

Explanation

Correct Answer

D:

Text recognition reads characters rather than merely describing visual objects. The expected acceptance evidence is: Handwritten serials are returned with confidence for review. This observation directly tests the stated behavior.

Incorrect Answers

A:

Image-and-text reasoning can connect a question with visible scene content. This observation does not establish the required result: Handwritten serials are returned with confidence for review.

B:

Different form elements require more than ordinary running-text OCR. This observation does not establish the required result: Handwritten serials are returned with confidence for review.

C:

Text-to-image generation creates new visual content from a description; classification only assigns labels to existing inputs. This observation does not establish the required result: Handwritten serials are returned with confidence for review.

Question 314

An Azure-based AI team must satisfy this objective: An inference service has high latency because every prompt includes a long irrelevant history. Which design choice is best?

  1. Represent text with embeddings and compare vectors using an appropriate similarity measure
  2. Explain token-based next-token prediction conditioned on preceding context
  3. Use an image-classification model with class probabilities
  4. Reduce unneeded context or summarize relevant prior turns

Explanation

Correct Answer

D:

Processing extra tokens increases cost and can affect response time. This implementation addresses the precise goal rather than an adjacent AI task.

Incorrect Answers

A:

Embeddings encode learned semantic relationships rather than literal string identity. That is not sufficient here: Processing extra tokens increases cost and can affect response time.

B:

A generative language model estimates likely continuations rather than reading a stored answer paragraph. That is not sufficient here: Processing extra tokens increases cost and can affect response time.

C:

Classification assigns image-wide categories rather than bounding-box coordinates. That is not sufficient here: Processing extra tokens increases cost and can affect response time.

Question 315

A project needs the following AI behavior: A Foundry app processes customer questions containing both a picture and text. What should the designer implement?

  1. Keep retrieved documents in lower-trust context and filter tool requests independently
  2. Send the image in a supported multimodal message along with the question
  3. Run adversarial model evaluations and verify that configured content-safety guardrails enforce the stated high-severity block criteria
  4. Add an approved knowledge/retrieval tool with clear description and access limits

Explanation

Correct Answer

B:

The model must receive the actual visual content to reason about it. This implementation addresses the precise goal rather than an adjacent AI task.

Incorrect Answers

A:

External content cannot replace privileged system behavior. That is not sufficient here: The model must receive the actual visual content to reason about it.

C:

Adversarial testing must measure actual configured policy enforcement rather than treating a prompt instruction as a guaranteed safety boundary. That is not sufficient here: The model must receive the actual visual content to reason about it.

D:

Tool descriptions guide selection, while backend rules enforce allowed sources. That is not sufficient here: The model must receive the actual visual content to reason about it.

Question 316

An Azure-based AI team must satisfy this objective: A spoken invoice number is regularly misrecognized as a similar-sounding sequence. Which design choice is best?

  1. Use speech-to-text recognition on a supported audio input stream
  2. Stop active recording and apply the documented microphone permission state
  3. Check synthesized voice availability for the selected locale and region
  4. Use digit-focused recognition hints and explicit confirmation for consequential IDs

Explanation

Correct Answer

D:

Speech recognition uncertainty is higher for confusable numeric strings. 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: Speech recognition uncertainty is higher for confusable numeric strings.

B:

User control of capture must be enforced at the device or application layer. That is not sufficient here: Speech recognition uncertainty is higher for confusable numeric strings.

C:

Not every voice is supported for all languages. That is not sufficient here: Speech recognition uncertainty is higher for confusable numeric strings.

Question 317

The approved design action is: Archive the actual evaluated response together with the prompt, model deployment, input and generation settings. Which test result would BEST establish that it worked?

  1. The summary references statements supported by the supplied passage
  2. An auditor can inspect the saved response alongside the exact prompt and model configuration used
  3. A valid identity with required resource role succeeds in the test
  4. The second tenant's test passes with its own authorized identity

Explanation

Correct Answer

B:

Recorded settings support investigation, but generative models can still vary on an apparently identical rerun. The expected acceptance evidence is: An auditor can inspect the saved response alongside the exact prompt and model configuration used. This observation directly tests the stated behavior.

Incorrect Answers

A:

Grounding constrains the answer to the available text and provenance. This observation does not establish the required result: An auditor can inspect the saved response alongside the exact prompt and model configuration used.

C:

Authentication proves caller identity and authorization grants resource access. This observation does not establish the required result: An auditor can inspect the saved response alongside the exact prompt and model configuration used.

D:

Copying code does not grant access to another tenant's resources. This observation does not establish the required result: An auditor can inspect the saved response alongside the exact prompt and model configuration used.

Question 318

An engineering review documents this requirement: A diagnostic agent runs a search but receives no matching documents. Select the most appropriate solution.

  1. Honor service throttling and use bounded exponential backoff
  2. Pass relevant conversation context or a verified summarized state within token limits
  3. Define required output sections and link citations to actual retrieved passages
  4. Implement a no-evidence response instead of fabricating a solution

Explanation

Correct Answer

D:

An empty result is not positive evidence for a generated factual claim. This implementation addresses the precise goal rather than an adjacent AI task.

Incorrect Answers

A:

Rate limits protect the service and uncontrolled retries amplify overload. That is not sufficient here: An empty result is not positive evidence for a generated factual claim.

B:

Stateful conversations require supplying or retrieving relevant prior context. That is not sufficient here: An empty result is not positive evidence for a generated factual claim.

C:

Explicit response structure makes evidence easier to verify. That is not sufficient here: An empty result is not positive evidence for a generated factual claim.

Question 319

A project needs the following AI behavior: A development team needs to experiment with a hosted model before integrating code. What should the designer implement?

  1. Require explicit confirmation and backend authorization before tool execution
  2. Create one agent in the Foundry portal with scoped tool definitions
  3. Define required output sections and link citations to actual retrieved passages
  4. Select and test a compatible deployed model through the Microsoft Foundry portal

Explanation

Correct Answer

D:

The portal provides model selection, deployment and test interfaces for supported resources. This implementation addresses the precise goal rather than an adjacent AI task.

Incorrect Answers

A:

High-impact side effects need permissions independent of conversational language. That is not sufficient here: The portal provides model selection, deployment and test interfaces for supported resources.

B:

A single agent can orchestrate a bounded tool without a multi-agent system. That is not sufficient here: The portal provides model selection, deployment and test interfaces for supported resources.

C:

Explicit response structure makes evidence easier to verify. That is not sufficient here: The portal provides model selection, deployment and test interfaces for supported resources.

Question 320

An Azure-based AI team must satisfy this objective: An insurance company wants to extract named policy limits from a contract. Which design choice is best?

  1. Create a schema for policy type, coverage amount and effective dates
  2. Enable a supported barcode extraction or recognition feature
  3. Use an appropriate Content Understanding or invoice-specific analyzer with a typed field schema
  4. Run supported image text extraction with field schema and source location

Explanation

Correct Answer

A:

Custom extraction maps semantically relevant phrases to typed fields. This implementation addresses the precise goal rather than an adjacent AI task.

Incorrect Answers

B:

A barcode is a structured visual symbol not ordinary OCR text. That is not sufficient here: Custom extraction maps semantically relevant phrases to typed fields.

C:

Document extraction turns unstructured pages into validated structured fields. That is not sufficient here: Custom extraction maps semantically relevant phrases to typed fields.

D:

Visual information extraction can map pixels to structured identifiers. That is not sufficient here: Custom extraction maps semantically relevant phrases to typed fields.

Microsoft Azure AI Fundamentals (AI-901) Practice Tests