Microsoft AI-102 (Designing and Implementing a Microsoft Azure AI Solution) Practice Test 60

Question 590

A document analyzer produces a category string that does not exist in the case-routing system's configured queues. What integration safeguard is needed?

  1. Validate the predicted category against a versioned routing map before assigning the case
  2. Represent timestamps and source identifiers consistently before merging multimodal observations
  3. Declare a supported categorical or enum-like field constraint for claim severity where available
  4. Verify supported input modality and the analyzer's declared processing configuration

Explanation

Correct Answer

A:

Classification outputs require mapping to authorized business queues; the predicted label alone should not be executed as an arbitrary destination. In the multimodal insurance-claim intake workflow, confirm the fix by retesting the affected behavior after validating the predicted category against a versioned routing map before assigning the case.

Incorrect Answers

B:

Although useful elsewhere in the multimodal insurance-claim intake workflow, a common event timeline needs normalized time representations and provenance, not simply concatenated text from each media type; the present case is different and calls for the team to validate the predicted category against a versioned routing map before assigning the case.

C:

That change addresses another failure mode: a typed or constrained field contract can restrict permissible categories; a free-form string invites inconsistent downstream values; it does not address this case, for which the relevant intervention is to validate the predicted category against a versioned routing map before assigning the case.

D:

Custom analyzer support is bounded by its defined content-type capabilities and version in its own workflow, but that is a different correction from the present requirement to validate the predicted category against a versioned routing map before assigning the case. The reported issue would remain.

Question 591

A team responsible for a multimodal insurance-claim intake workflow needs to decide whether uncertain or poorly supported extracted fields need human verification; the acceptance check is that low-confidence or unsupported values are routed for review. Which capability should be selected?

  1. Configure a custom Content Understanding analyzer
  2. Use Content Understanding document classification
  3. Inspect analyzer confidence and extraction evidence
  4. Define an analyzer field schema

Explanation

Correct Answer

C:

Validation of structured outputs should consider confidence or evidence where available and compare results to source content. For this multimodal insurance-claim intake workflow, the intended evidence is that low-confidence or unsupported values are routed for review. A JSON field existing in the response is not proof that its inferred value is correct.

Incorrect Answers

A:

Content Understanding analyzers declare supported content processing and extraction settings for the target modality; it may be used to define the processing behavior for a specified content type and its desired outputs in a multimodal insurance-claim intake workflow, but the current requirement is instead to decide whether uncertain or poorly supported extracted fields need human verification. That separate capability does not prove that low-confidence or unsupported values are routed for review; Creating only a general search index does not define an analyzer's extraction behavior, so selecting that alternative would not meet the stated acceptance check.

B:

A classifier-oriented analyzer can map document content into configured categories to support routing or review; it may be used to categorize submitted documents by their semantic or business type in a multimodal insurance-claim intake workflow, but the current requirement is instead to decide whether uncertain or poorly supported extracted fields need human verification. That separate capability does not prove that low-confidence or unsupported values are routed for review; Classification does not by itself transcribe a phone call into words, so selecting that alternative would not meet the stated acceptance check.

D:

A field schema guides the analyzer to emit application-ready values such as policy number, category, or summary; it may be used to specify named output attributes and their data types for structured extraction in a multimodal insurance-claim intake workflow, but the current requirement is instead to decide whether uncertain or poorly supported extracted fields need human verification. That separate capability does not prove that low-confidence or unsupported values are routed for review; Unconstrained plain OCR output does not guarantee those typed business fields, so selecting that alternative would not meet the stated acceptance check.

Question 592

A claim combines video, a photo, and narration, and extracted facts refer to different times. What normalization is required?

  1. Validate the predicted category against a versioned routing map before assigning the case
  2. Route the field for manual verification against the source record
  3. Represent timestamps and source identifiers consistently before merging multimodal observations
  4. Verify supported input modality and the analyzer's declared processing configuration

Explanation

Correct Answer

C:

A common event timeline needs normalized time representations and provenance, not simply concatenated text from each media type. In the multimodal insurance-claim intake workflow, confirm the fix by retesting the affected behavior after representing timestamps and source identifiers consistently before merging multimodal observations.

Incorrect Answers

A:

Classification outputs require mapping to authorized business queues; the predicted label alone should not be executed as an arbitrary destination in its own workflow, but that is a different correction from the present requirement to represent timestamps and source identifiers consistently before merging multimodal observations. The reported issue would remain.

B:

That change addresses another failure mode: existence of a JSON value is not proof of accuracy, particularly without reliable supporting evidence; it does not address this case, for which the relevant intervention is to represent timestamps and source identifiers consistently before merging multimodal observations.

D:

Although useful elsewhere in the multimodal insurance-claim intake workflow, custom analyzer support is bounded by its defined content-type capabilities and version; the present case is different and calls for the team to represent timestamps and source identifiers consistently before merging multimodal observations.

Question 593

An application rejects every claim when a model omits an optional extracted field. What output validation rule is safer?

  1. Distinguish required, missing, and uncertain fields according to the business schema
  2. Retain media time offsets or source identifiers for extracted observations where available
  3. Define a supported repeated item or array field with explicit item structure
  4. Address the intended analyzer ID and supported version in the processing request

Explanation

Correct Answer

A:

Missing optional information should not automatically invalidate an entire claim; validation should reflect field criticality. In the multimodal insurance-claim intake workflow, confirm the fix by retesting the affected behavior after distinguishing required, missing, and uncertain fields according to the business schema.

Incorrect Answers

B:

Although useful elsewhere in the multimodal insurance-claim intake workflow, temporal provenance helps reviewers compare assertions with exact portions of recordings; the present case is different and calls for the team to distinguish required, missing, and uncertain fields according to the business schema.

C:

Repeated structured content requires an appropriate collection type, not a free-form string in its own workflow, but that is a different correction from the present requirement to distinguish required, missing, and uncertain fields according to the business schema. The reported issue would remain.

D:

That change addresses another failure mode: an analyzer definition cannot influence runtime if clients invoke another resource configuration; it does not address this case, for which the relevant intervention is to distinguish required, missing, and uncertain fields according to the business schema.

Question 594

A classifier sees a mixed multipage packet with both an invoice and a letter. What should the input design consider?

  1. Retain media time offsets or source identifiers for extracted observations where available
  2. Define a supported repeated item or array field with explicit item structure
  3. Distinguish required, missing, and uncertain fields according to the business schema
  4. Use supported document segmentation or per-document classification rather than one assumed packet label

Explanation

Correct Answer

D:

Mixed packets may contain several categories; applying one category to all pages loses routing detail. In the multimodal insurance-claim intake workflow, confirm the fix by retesting the affected behavior after using supported document segmentation or per-document classification rather than one assumed packet label.

Incorrect Answers

A:

Although useful elsewhere in the multimodal insurance-claim intake workflow, temporal provenance helps reviewers compare assertions with exact portions of recordings; the present case is different and calls for the team to use supported document segmentation or per-document classification rather than one assumed packet label.

B:

Repeated structured content requires an appropriate collection type, not a free-form string in its own workflow, but that is a different correction from the present requirement to use supported document segmentation or per-document classification rather than one assumed packet label. The reported issue would remain.

C:

That change addresses another failure mode: missing optional information should not automatically invalidate an entire claim; validation should reflect field criticality; it does not address this case, for which the relevant intervention is to use supported document segmentation or per-document classification rather than one assumed packet label.

Question 595

Two media sources disagree about the reported incident date. What should the aggregator do?

  1. Compare extracted field values and grounding against a labeled held-out reference set
  2. Regression-test the analyzer definition against versioned representative input files
  3. Flag conflicting source-grounded fields instead of silently choosing one generated value
  4. Treat classification as document type detection and separately validate extracted monetary fields

Explanation

Correct Answer

C:

Multimodal extraction can surface inconsistent evidence; resolution should be auditable and policy-driven. In the multimodal insurance-claim intake workflow, confirm the fix by retesting the affected behavior after flaging conflicting source-grounded fields instead of silently choosing one generated value.

Incorrect Answers

A:

Although useful elsewhere in the multimodal insurance-claim intake workflow, response-format success does not measure semantic field correctness or evidence faithfulness; the present case is different and calls for the team to flag conflicting source-grounded fields instead of silently choosing one generated value.

B:

Analyzer configuration and prompts affect extraction behavior and should be evaluated before promotion in its own workflow, but that is a different correction from the present requirement to flag conflicting source-grounded fields instead of silently choosing one generated value. The reported issue would remain.

D:

That change addresses another failure mode: document category prediction does not certify the accuracy of underlying prices or claims; it does not address this case, for which the relevant intervention is to flag conflicting source-grounded fields instead of silently choosing one generated value.

Question 596

A claims team changes its extraction instructions and confidence drops on sample claims. What release practice helps?

  1. Version the analyzer field schema and coordinate changes with consumers
  2. Compare extracted field values and grounding against a labeled held-out reference set
  3. Regression-test the analyzer definition against versioned representative input files
  4. Flag conflicting source-grounded fields instead of silently choosing one generated value

Explanation

Correct Answer

C:

Analyzer configuration and prompts affect extraction behavior and should be evaluated before promotion. In the multimodal insurance-claim intake workflow, confirm the fix by retesting the affected behavior after regression-testing the analyzer definition against versioned representative input files.

Incorrect Answers

A:

That change addresses another failure mode: changing a field name is an API-contract change even when source understanding remains accurate; it does not address this case, for which the relevant intervention is to regression-test the analyzer definition against versioned representative input files.

B:

Although useful elsewhere in the multimodal insurance-claim intake workflow, response-format success does not measure semantic field correctness or evidence faithfulness; the present case is different and calls for the team to regression-test the analyzer definition against versioned representative input files.

D:

Multimodal extraction can surface inconsistent evidence; resolution should be auditable and policy-driven in its own workflow, but that is a different correction from the present requirement to regression-test the analyzer definition against versioned representative input files. The reported issue would remain.

Question 597

A document is classified as “repair estimate,” and staff assume all quoted costs have been verified. What distinction should be enforced?

  1. Compare extracted field values and grounding against a labeled held-out reference set
  2. Treat classification as document type detection and separately validate extracted monetary fields
  3. Flag conflicting source-grounded fields instead of silently choosing one generated value
  4. Regression-test the analyzer definition against versioned representative input files

Explanation

Correct Answer

B:

Document category prediction does not certify the accuracy of underlying prices or claims. In the multimodal insurance-claim intake workflow, confirm the fix by retesting the affected behavior after treating classification as document type detection and separately validate extracted monetary fields.

Incorrect Answers

A:

That change addresses another failure mode: response-format success does not measure semantic field correctness or evidence faithfulness; it does not address this case, for which the relevant intervention is to treat classification as document type detection and separately validate extracted monetary fields.

C:

Multimodal extraction can surface inconsistent evidence; resolution should be auditable and policy-driven in its own workflow, but that is a different correction from the present requirement to treat classification as document type detection and separately validate extracted monetary fields. The reported issue would remain.

D:

Although useful elsewhere in the multimodal insurance-claim intake workflow, analyzer configuration and prompts affect extraction behavior and should be evaluated before promotion; the present case is different and calls for the team to treat classification as document type detection and separately validate extracted monetary fields.

Question 598

A claims extractor outputs arbitrary strings for severity but the business accepts only `minor`, `moderate`, or `major`. How should the schema constrain this field?

  1. Declare a supported categorical or enum-like field constraint for claim severity where available
  2. Validate the predicted category against a versioned routing map before assigning the case
  3. Verify supported input modality and the analyzer's declared processing configuration
  4. Route the field for manual verification against the source record

Explanation

Correct Answer

A:

A typed or constrained field contract can restrict permissible categories; a free-form string invites inconsistent downstream values. In the multimodal insurance-claim intake workflow, confirm the fix by retesting the affected behavior after declaring a supported categorical or enum-like field constraint for claim severity where available.

Incorrect Answers

B:

That change addresses another failure mode: classification outputs require mapping to authorized business queues; the predicted label alone should not be executed as an arbitrary destination; it does not address this case, for which the relevant intervention is to declare a supported categorical or enum-like field constraint for claim severity where available.

C:

Custom analyzer support is bounded by its defined content-type capabilities and version in its own workflow, but that is a different correction from the present requirement to declare a supported categorical or enum-like field constraint for claim severity where available. The reported issue would remain.

D:

Although useful elsewhere in the multimodal insurance-claim intake workflow, existence of a JSON value is not proof of accuracy, particularly without reliable supporting evidence; the present case is different and calls for the team to declare a supported categorical or enum-like field constraint for claim severity where available.

Question 599

The solution brief for a multimodal insurance-claim intake workflow requires the system to analyze supported image, video, audio, and document inputs in a unified extraction workflow. A successful test confirms that different supported media produce consistent claim-oriented output fields. What should the team use?

  1. Configure a custom Content Understanding analyzer
  2. Use multimodal Content Understanding processing
  3. Inspect analyzer confidence and extraction evidence
  4. Define an analyzer field schema

Explanation

Correct Answer

B:

Multimodal analyzers can derive structured fields and relevant content insights from supported media types. For this multimodal insurance-claim intake workflow, the intended evidence is that different supported media produce consistent claim-oriented output fields. A receipt-only prebuilt extractor cannot be assumed to process a recorded claim interview.

Incorrect Answers

A:

Content Understanding analyzers declare supported content processing and extraction settings for the target modality; it may be used to define the processing behavior for a specified content type and its desired outputs in a multimodal insurance-claim intake workflow, but the current requirement is instead to analyze supported image, video, audio, and document inputs in a unified extraction workflow. That separate capability does not prove that different supported media produce consistent claim-oriented output fields; Creating only a general search index does not define an analyzer's extraction behavior, so selecting that alternative would not meet the stated acceptance check.

C:

Validation of structured outputs should consider confidence or evidence where available and compare results to source content; it may be used to decide whether uncertain or poorly supported extracted fields need human verification in a multimodal insurance-claim intake workflow, but the current requirement is instead to analyze supported image, video, audio, and document inputs in a unified extraction workflow. That separate capability does not prove that different supported media produce consistent claim-oriented output fields; A JSON field existing in the response is not proof that its inferred value is correct, so selecting that alternative would not meet the stated acceptance check.

D:

A field schema guides the analyzer to emit application-ready values such as policy number, category, or summary; it may be used to specify named output attributes and their data types for structured extraction in a multimodal insurance-claim intake workflow, but the current requirement is instead to analyze supported image, video, audio, and document inputs in a unified extraction workflow. That separate capability does not prove that different supported media produce consistent claim-oriented output fields; Unconstrained plain OCR output does not guarantee those typed business fields, so selecting that alternative would not meet the stated acceptance check.

Question 600

A downstream parser breaks when analysts rename the output field `policyId`. What contract process is missing?

  1. Flag conflicting source-grounded fields instead of silently choosing one generated value
  2. Treat classification as document type detection and separately validate extracted monetary fields
  3. Version the analyzer field schema and coordinate changes with consumers
  4. Regression-test the analyzer definition against versioned representative input files

Explanation

Correct Answer

C:

Changing a field name is an API-contract change even when source understanding remains accurate. In the multimodal insurance-claim intake workflow, confirm the fix by retesting the affected behavior after versioning the analyzer field schema and coordinate changes with consumers.

Incorrect Answers

A:

Multimodal extraction can surface inconsistent evidence; resolution should be auditable and policy-driven in its own workflow, but that is a different correction from the present requirement to version the analyzer field schema and coordinate changes with consumers. The reported issue would remain.

B:

That change addresses another failure mode: document category prediction does not certify the accuracy of underlying prices or claims; it does not address this case, for which the relevant intervention is to version the analyzer field schema and coordinate changes with consumers.

D:

Although useful elsewhere in the multimodal insurance-claim intake workflow, analyzer configuration and prompts affect extraction behavior and should be evaluated before promotion; the present case is different and calls for the team to version the analyzer field schema and coordinate changes with consumers.

Designing and Implementing a Microsoft Azure AI Solution (AI-102) Practice Tests