cert
cert-1
cert-2

Pass Microsoft AI-103 Exam in First Attempt Guaranteed!

Get 100% Latest Exam Questions, Accurate & Verified Answers to Pass the Actual Exam!
30 Days Free Updates, Instant Download!

cert-5
cert-6
AI-103 Exam - Verified By Experts
AI-103 Premium File

AI-103 Premium File

$89.99
$98.99
  • Premium File 187 Questions & Answers. Last Update: Nov 15, 2026

Whats Included:

  • Latest Questions
  • 100% Accurate Answers
  • Fast Exam Updates
 
$98.99
$89.99
accept 397 downloads in the last 7 days
block-screenshots
Microsoft AI-103 Practice Test 12

Question 110

An ingestion team wants to minimize exposing raw call recordings to broad users. What security measure fits?

  1. Grant search users direct access to every underlying call recording after granting transcript access
  2. Make the entire transcript and source audio accessible to every employee so search results never fail a permission check.
  3. Store speaker or channel attribution with transcript spans
  4. Enforce document-level access rules on transcript and audio source references

Explanation

Correct Answer

D:

Audio-derived transcripts can contain sensitive information and require source-equivalent permissions. The practical requirement is to apply source ACLs to audio transcripts; a solution built to navigate from text hits to audio evidence would not meet that same need.

Incorrect Answers

A:

A transcript access grant should not silently confer access to sensitive source audio. Retrieval and playback must separately enforce the authorized document and recording permissions.

B:

Search convenience cannot override customer recording privacy. Enforce document-level authorization for both indexed transcript snippets and linked audio.

C:

Speaker identity is essential when retrieving who said a particular phrase. The function serves to separate speakers in indexed audio, whereas the actual requirement is to apply source ACLs to audio transcripts.

Question 111

An agent receives a successful tool result and then a model response indicating normal completion. When should the loop stop?

  1. Continue the orchestration loop after a successful completion signal solely because unused tool-call slots remain in the budget.
  2. Launch extra tool calls in parallel after the model has returned a valid final completion.
  3. Terminate on the actual completion signal after required tool results have been applied
  4. Wait for a nonexistent refund result after the assistant has already received the tool result and issued a final answer.

Explanation

Correct Answer

C:

The stop reason, not the presence of some text, distinguishes completion from an intermediate tool request. This directly supports the need to terminate a tool loop only after a completed response, unlike a mechanism primarily intended to stop runaway tool usage.

Incorrect Answers

A:

The controller should stop once required tool results are incorporated and the model has actually completed. Resource budgets are upper limits, not reasons to make more calls.

B:

More tool invocations after terminal completion create unnecessary side effects and cost. The orchestration loop should stop after an authorized final answer with no pending required calls.

D:

Terminal completion with all required tool outputs processed should end the loop. Inventing another dependency risks deadlock or redundant execution.

Question 112

A travel assistant needs to perform a reservation update rather than merely describe the booking policy. What must its design include?

  1. Have the model describe how reservation updates generally work without providing an authorized booking-change operation.
  2. Remember a booking-change request in session state but provide no authorized reservation-update API to execute it.
  3. Expose a narrowly permissioned action tool with explicit input validation
  4. Store explicit durable preferences in a governed profile store, excluding secrets

Explanation

Correct Answer

C:

An external side effect requires an authorized tool boundary; model text alone cannot update the booking system. The decisive difference is whether the design can perform an approved business action through a tool; the competing function serves to filter retrieved knowledge by the user's permission.

Incorrect Answers

A:

A descriptive answer does not make the requested modification. An authenticated, narrowly permissioned action tool must validate and execute approved changes.

B:

State retention is not an action mechanism. An authenticated, validated write tool is required to change a booking rather than merely explain the policy.

D:

Long-lived user preferences need a controlled source with consent and deletion policies rather than raw unbounded chat logs. That is relevant to persist allowed user preferences across sessions, but the present scenario requires an ability to perform an approved business action through a tool.

Question 113

A designer needs to replace one product logo in a short generated video while maintaining temporal consistency. What workflow is needed?

  1. Run one independent still-image edit for each video frame without tracking the logo region across time or checking temporal consistency.
  2. Use a supported frame/region-based video edit with temporal consistency checks
  3. Replace the logo in one isolated frame and leave all adjacent video frames untracked.
  4. Evaluate preservation of unmasked regions as a specific acceptance criterion

Explanation

Correct Answer

B:

Video modifications must remain coherent across frames; isolated frame edits often flicker. This directly supports the need to edit video content coherently across time, unlike a mechanism primarily intended to extend the image beyond its existing frame.

Incorrect Answers

A:

Uncoordinated frame edits can make the logo flicker or drift. A supported video editing workflow must preserve region tracking and continuity across frames.

C:

A single-frame substitution may flicker or drift as the clip moves. Temporal tracking or consistency-aware video editing is needed across frames.

D:

Localized edits should be tested for unintended changes outside the approved region. Its primary purpose is to check image preservation outside a target mask; the stated constraint instead demands a way to edit video content coherently across time.

Question 114

A retailer wants to measure whether customer feedback is positive, neutral or negative. Which text output is appropriate?

  1. Count the number of words in each review and label long reviews positive without analyzing the customer's language.
  2. Classify sentiment polarity from the customer's feedback
  3. Evaluate communicative intent separately from surface tone
  4. Use contextual examples and evaluate sarcasm-specific errors

Explanation

Correct Answer

B:

Sentiment captures evaluative polarity rather than merely detecting subject keywords. The decisive difference is whether the design can measure positive, negative or neutral opinion; the competing function serves to identify mixed sentiment about separate product aspects.

Incorrect Answers

A:

Review length does not determine whether a customer expresses praise, dissatisfaction or neutrality. A sentiment classifier must inspect relevant wording and context.

C:

Polite tone and escalatory intent can coexist, so one polarity label is insufficient. Its primary purpose is to separate message intent from emotional tone; the stated constraint instead demands a way to measure positive, negative or neutral opinion.

D:

Literal polarity words can contradict contextual meaning, so domain testing is necessary. That is relevant to recognize sentiment polarity in sarcastic context, but the present scenario requires an ability to measure positive, negative or neutral opinion.

Question 115

A media archive needs natural-language search for events within long training videos. Which preprocessing pipeline helps?

  1. Index only one generic embedding for the video's filename and omit timestamps, scenes and frame descriptions.
  2. Extract timestamped scenes or frames and generate indexed visual descriptors
  3. Associate each scene hit with video ID and precise time range
  4. Preserve modality, asset and temporal relationships across indexed segments

Explanation

Correct Answer

B:

Segment-based visual metadata makes otherwise opaque video content searchable. Here the relevant capability must let the team create searchable video-scene metadata. An alternative targeting search image collections by visual similarity solves another problem.

Incorrect Answers

A:

A filename-level embedding loses what appears within individual moments of a long recording. Searchable event retrieval needs time-linked visual or transcript evidence for the relevant scene.

C:

Timestamped source links let reviewers validate a visual-search claim. That capability is appropriate for link scene results to original footage, not this separate task of finding how to create searchable video-scene metadata.

D:

Typed links prevent text from one source segment being misattributed to another. Here the needed outcome is to create searchable video-scene metadata; an option designed to maintain multimodal source associations would not directly provide it.

Question 116

An orchestration controller runs too many repeated tool calls while making no progress. What resource guard should be enforced?

  1. Retry the same failing external tool indefinitely, because a structured error result makes every retry safe.
  2. Reject invalid tool arguments and request bounded correction before any side effect
  3. Use a stable idempotency identifier and reconcile the existing invoice before retry
  4. Limit tool iterations and elapsed time and detect repeated no-progress calls

Explanation

Correct Answer

D:

An external execution budget keeps autonomy bounded even if prompts would otherwise loop. The decisive difference is whether the design can stop runaway tool usage; the competing function serves to report tool failure instead of fabricated emptiness.

Incorrect Answers

A:

A structured error is useful for diagnosis but does not cap resource consumption. Without iteration, elapsed-time and no-progress limits, an agent can repeatedly call tools without advancing the task.

B:

A model tool call must satisfy deterministic schema validation and business rules before a sensitive write. Here the needed outcome is to stop runaway tool usage; an option designed to block malformed tool write arguments would not directly provide it.

C:

Uncertain completion can duplicate financial side effects; idempotency and read-before-retry reduce that risk. Its primary purpose is to avoid duplicate writes after timeout; the stated constraint instead demands a way to stop runaway tool usage.

Question 117

A service needs predictable deployment promotion from testing to production without silently adopting a newly published model revision. What should be recorded?

  1. Deploy whichever model is currently the newest in the catalog.
  2. Retain only the model display name and discard its version and configuration.
  3. Pin the approved model version and deployment configuration in version-controlled release artifacts.
  4. Reuse a staging endpoint in production without release metadata.

Explanation

Correct Answer

C:

An explicitly approved version avoids unnoticed behavior changes between test and production; promotion should include regression evaluations and rollback metadata.

Incorrect Answers

A:

Automatic version drift can change behavior after tests without an explicit promotion decision.

B:

A friendly name is insufficient to reproduce the previously validated serving behavior.

D:

Sharing endpoints does not establish that the exact tested model revision and settings are deployed.

Question 118

An application makes a masked edit, but reviewers cannot identify which source asset was used. What should be archived?

  1. Retain source media, mask versions, edit prompt and output version IDs
  2. Run another masked edit and keep only the resulting image, without storing the source image or mask revision.
  3. Verify frame-to-frame video smoothness without recording the source image, edit mask or generation identifiers.
  4. Archive the finished edited picture alone, deleting the original image, mask, prompt and version metadata.

Explanation

Correct Answer

A:

Versioning masks and source assets makes region editing auditable and reproducible. The decisive difference is whether the design can reproduce a localized image edit; the competing function serves to replace only a selected image region.

Incorrect Answers

B:

Re-running inpainting does not reconstruct the original source and editable-region evidence. Auditability depends on retaining source media identifiers, mask versions, prompt and output provenance.

C:

Temporal smoothness does not establish which original asset and mask produced a still-image edit. Record source, mask, prompt and output versions for auditability.

D:

Without the source, edit region and prompt lineage, reviewers cannot reproduce which pixels were intentionally changed. Retain the relevant artifacts and IDs.

Question 119

An email is politely worded but threatens legal action. Which assessment should distinguish style from intent?

  1. Mark the email as benign solely because its wording has positive sentiment
  2. Evaluate communicative intent separately from surface tone
  3. Classify the message as positive because it uses polite language, without checking the legal threat in its content.
  4. Score sentiment at the aspect or topic level rather than only document level

Explanation

Correct Answer

B:

Polite tone and escalatory intent can coexist, so one polarity label is insufficient. In this scenario the team must separate message intent from emotional tone. A feature used to recognize sentiment polarity in sarcastic context does not substitute for it.

Incorrect Answers

A:

Positive surface tone does not rule out a threat of litigation or a demand for action. Communicative intent must be evaluated separately from overall sentiment polarity.

C:

Polite style can coexist with threatening intent. Analyze communicative purpose and risk separately from the sentiment of surface wording.

D:

Aspect sentiment avoids collapsing mixed reactions into one misleading global score. This helps when a team needs to identify mixed sentiment about separate product aspects; it does not establish how to separate message intent from emotional tone.

Question 120

A retrieved video snippet must be reviewed in the original file. What provenance should the index store?

  1. Save only a generated clip caption without the original video asset ID or time interval
  2. Associate each scene hit with video ID and precise time range
  3. Store the scene's textual caption without a video identifier or start and end time markers linking it to the original recording.
  4. Extract timestamped scenes or frames and generate indexed visual descriptors

Explanation

Correct Answer

B:

Timestamped source links let reviewers validate a visual-search claim. This directly supports the need to link scene results to original footage, unlike a mechanism primarily intended to search image collections by visual similarity.

Incorrect Answers

A:

A caption is not a reliable pointer to the evidence location. Video retrieval results need an asset identifier and exact timestamps so a reviewer can replay the claimed event.

C:

A caption alone cannot navigate a reviewer to the precise evidence. The index needs the source video ID and time range for each retrieved scene.

D:

Segment-based visual metadata makes otherwise opaque video content searchable. Its primary purpose is to create searchable video-scene metadata; the stated constraint instead demands a way to link scene results to original footage.

Developing AI Apps and Agents on Azure (AI-103) Practice Tests