McAfee Secured

Exam Code: AI-103

Exam Name: Developing AI Apps and Agents on Azure

Certification Provider: Microsoft

Corresponding Certification: Microsoft Certified: Azure AI Apps and Agents Developer Associate

100% Microsoft AI-103 Guaranteed Success With Testking Training Materials

Get Real AI-103 Exam Questions & Expert Verified Answers as Experienced on the Real Exam!

187 Questions & Answers with Testing Engine

"Developing AI Apps and Agents on Azure Exam", also known as AI-103 exam, is a Microsoft certification exam.

Pass your tests with the always up-to-date AI-103 Exam Engine. Your AI-103 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-103 Practice Test 23

Question 220

A source assistant has two tools: `search_policy` and `update_policy`. The user asks what the current rule says. What choice is proper?

  1. Create a new policy document automatically
  2. Invoke only the read-only policy lookup tool
  3. Invoke both tools to be thorough
  4. Update the policy to match the user's question

Explanation

Correct Answer

B:

An informational request does not authorize changing official policy.

Incorrect Answers

A:

Content creation is not warranted by the lookup task.

C:

Unnecessary writes could alter current company guidance.

D:

Reading a rule is not a request to modify it.

Question 221

A team needs repeatable Foundry deployments through Terraform or Bicep. What infrastructure practice matters?

  1. Store all resource secrets as public template outputs
  2. Configure production only through undocumented portal clicks
  3. Version deployment settings, identities and connections as governed code
  4. Delete the production resource on every pipeline execution

Explanation

Correct Answer

C:

Repeatable templates reduce configuration drift across approved environments.

Incorrect Answers

A:

Deployment source must not disclose credentials to consumers.

B:

Unrecorded mutations cannot be reproduced reliably by CI.

D:

Destructive recreation is unnecessary for normal desired-state updates.

Question 222

An agent's response repeatedly cites the same document for unrelated products. What groundedness check helps?

  1. Increase inference token budget to add citations automatically
  2. Only count the number of URLs printed
  3. Verify product ID and supporting source match for each claim
  4. Use one popular document as a global reference

Explanation

Correct Answer

C:

Evidence must be both relevant to the requested entity and substantively supportive.

Incorrect Answers

A:

Longer outputs may multiply unsupported references.

B:

Many links can still point to unrelated materials.

D:

Broad citation reuse does not prove each product claim.

Question 223

An image-search assistant mixes up two product photos with the same filename. What source record should it use?

  1. Assign all assets one ID for convenience
  2. Stable image IDs and hashes with per-asset feature metadata
  3. Use only the upload time to interpret product features
  4. Assume matching names prove images are identical

Explanation

Correct Answer

B:

File names can collide; content-linked identity supports correct result attribution.

Incorrect Answers

A:

Collapsing identity prevents per-image provenance.

C:

Time does not describe which product is pictured.

D:

Two assets can share a filename with different pixels.

Question 224

An app uses two models: a small one for routing, a large one for hard reasoning. How should routing be tuned?

  1. Route everything to the smaller model regardless of failure rates
  2. Route everything to the larger model without measuring gains
  3. Define confidence/complexity thresholds and evaluate misroutes on representative cases
  4. Switch models based only on current user message length

Explanation

Correct Answer

C:

Cost-efficient escalation requires protecting difficult requests from wrong lower-tier selection.

Incorrect Answers

A:

Low cost cannot justify unacceptable task errors.

B:

Universal escalation may waste resources on trivial cases.

D:

Short questions can still require complex reasoning or tools.

Question 225

A service is migrated to a different Foundry project and private network. Which TWO areas should be revalidated?

  1. Project endpoint, deployment identifier and workload identity permissions
  2. Increase model response temperature
  3. Private endpoint name resolution and effective application routes
  4. Only update the model version without validating project identity and network routing
  5. Delete source URLs from the knowledge index

Explanation

Correct Answers

A:

The client must address and authenticate to the newly approved target.

C:

A correct service identity alone cannot fix unreachable private network endpoints.

Incorrect Answers

B:

Randomness does not solve a wrong endpoint or private DNS.

D:

Model version changes cannot correct an invalid private endpoint or missing identity permissions.

E:

Source removal does not provide connectivity.

Question 226

A research agent needs only the section on retention from a 200-page PDF. What retrieval behavior is efficient?

  1. Insert all 200 pages into every prompt
  2. Retrieve the relevant section with context and exact page citation
  3. Read only the document cover and guess the section
  4. Use an unrelated summary from another vendor

Explanation

Correct Answer

B:

Targeted chunks reduce prompt cost while preserving source support.

Incorrect Answers

A:

Excess context increases latency and distracts generation.

C:

The title page cannot provide a detailed retention rule.

D:

Different documents may have incompatible policies.

Question 227

A team wants to reason about both a customer's spoken complaint and an accompanying sound of machinery. What model pipeline suits?

  1. Only ordinary speech-to-text with no acoustic analysis
  2. Only image generation of a machine
  3. Audio-aware multimodal interpretation of speech and non-speech acoustic cues
  4. Only sentiment classification of a text caption

Explanation

Correct Answer

C:

A transcript may omit mechanical sounds that explain the problem.

Incorrect Answers

A:

Nonverbal sound evidence would be discarded.

B:

Synthetic images cannot identify the actual recording's sound.

D:

Text tone does not describe the acoustic malfunction.

Question 228

A document analyzer extracts a total but no indication of which page it came from. What should output include?

  1. Field value with source spans, page location and confidence when supported
  2. Return the document ID and field value, but omit page and span references
  3. Return the last page number as the location for every amount
  4. Return a confidence score with no source bounding region

Explanation

Correct Answer

A:

Location-level grounding allows reviewers to verify extracted facts.

Incorrect Answers

B:

File-level provenance cannot locate the particular total on a multi-page document.

C:

A constant page reference will be misleading whenever the total appears elsewhere.

D:

Confidence does not identify where a reviewer can inspect the extracted value.

Question 229

A marketing image editor must remove one object while retaining exact product geometry. Which TWO elements should the edit request contain?

  1. The original approved reference image and a localized edit mask
  2. Use only a text sentiment model to evaluate pixel preservation
  3. Explicit preservation constraints and a comparison of protected regions
  4. Generate a new unrelated product from text only
  5. Delete source imagery before checking the result

Explanation

Correct Answers

A:

The mask identifies pixels eligible for transformation.

C:

Output review detects changes outside the requested region.

Incorrect Answers

B:

Sentiment cannot inspect modified image geometry.

D:

Unconditioned synthesis can change the depicted merchandise.

E:

Without the reference, fidelity cannot be verified.

Question 230

A document retriever produces overlapping duplicate chunks from the same page. What should ranking consider?

  1. Remove all source IDs before deduplication
  2. Add more duplicates to strengthen apparent certainty
  3. Return only the longest chunk from every document
  4. Deduplicate near-identical spans while retaining source coverage and context

Explanation

Correct Answer

D:

Repeated excerpts should not crowd out other relevant evidence.

Incorrect Answers

A:

Without identity, overlap becomes harder to identify correctly.

B:

Repetition is not independent corroboration.

C:

Length does not establish relevance or unique evidence.

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