Which solutions meet regional loan-assessment GenAI requirements on Bedrock?

AIP-C01 By islever Editorial Last updated

Design and implement enterprise integration architectures. Implement data security and privacy controls. Implement AI governance and compliance mechanisms.

Short answer

ABE Use Textract with A2I in-Region for extraction and low-confidence review, Lambda to redact PII before Bedrock, and Step Functions to orchestrate prompts.

Question

A bank is building a generative AI (GenAI) application that uses Amazon Bedrock to assess loan applications by using scanned financial documents. The application must extract structured data from the documents. The application must redact personally identifiable information (PII) before inference. The application must use foundation models (FMs) to generate approvals. The application must route low-confidence document extraction results to human reviewers who are within the same AWS Region as the loan applicant.
The company must ensure that the application complies with strict Regional data residency and auditability requirements. The application must be able to scale to handle 25,000 applications each day and provide 99.9% availability.
Which combination of solutions will meet these requirements? (Choose three.)
  1. Deploy Amazon Textract and Amazon Augmented AI (Amazon A2I) within the same Region to extract relevant data from the scanned documents. Route low-confidence pages to human reviewers. Correct Answer
  2. Use AWS Lambda functions to detect and redact PII from submitted documents before inference. Apply Amazon Bedrock guardrails to prevent inappropriate or unauthorized content in model outputs. Configure Region-specific IAM roles to enforce data residency requirements and to control access to the extracted data. Correct Answer
  3. Use Amazon Kendra and Amazon OpenSearch Service to extract field level values semantically from the uploaded documents before inference.
  4. Store uploaded documents in Amazon S3 and apply object metadata. Configure IAM policies to store original documents within the same Region as each applicant. Enable object tagging for future audits.
  5. Use AWS Glue Data Quality to validate the structured document data. Use AWS Step Functions to orchestrate a review workflow that includes a prompt engineering step that transforms validated data into optimized prompts before invoking Amazon Bedrock to assess loan applications. Correct Answer
  6. Use Amazon SageMaker Clarify to generate fairness and bias reports based on model scoring decisions that Amazon Bedrock makes.

Key Takeaways

The question tests whether you can map each mandatory control—Textract/A2I extraction and in-Region human review, Lambda and guardrails PII redaction before Bedrock, and Step Functions orchestration—to a managed AWS service, while the trap is choosing S3 tagging in D as if storage alone proved residency and auditability.

This AIP-C01 scenario tests a multi-service Bedrock loan-assessment pipeline that must extract scanned documents, redact PII before inference, and route low-confidence work to in-Region reviewers. The page establishes why Amazon Textract with A2I, Lambda-based redaction plus guardrails, and Step Functions orchestration (A, B, E) satisfy the residency, audit, scale, and availability requirements.

Common mistake: The most common wrong answer is ABD, because same-Region S3 storage and object tagging sound like they satisfy residency and auditability, but D does not redact PII before inference or orchestrate the Bedrock loan-assessment workflow that E provides.

Expert Analysis

Why the Answer Is Correct

A is mandatory because Amazon Textract extracts structured fields from scanned financial documents, and Amazon A2I provides the managed human-review loop for low-confidence pages; deploying both in the applicant's Region supports the residency requirement. B is mandatory because Lambda performs PII detection and redaction before any Bedrock inference, Bedrock guardrails add output controls, and Region-scoped IAM roles enforce access boundaries. E supplies the orchestration layer: Step Functions coordinates extraction, validation, prompt construction, and the Bedrock loan-assessment call while retaining execution history for audit, and Glue Data Quality validates the structured fields before they are used. Together A, B, and E cover extraction, pre-inference redaction, in-Region human review, FM approval generation, residency and audit controls, and managed scaling for 25,000 daily applications with 99.9% availability.

Why the Other Options Are Wrong

C is wrong because Amazon Kendra and OpenSearch are retrieval/search services; they do not perform OCR or field-level extraction from scanned financial documents. D is the strongest distractor: S3 object metadata, Region-specific buckets, and tagging help audit and residency, but they neither redact PII before inference nor provide the Step Functions prompt orchestration that invokes Bedrock for approvals, and IAM policies alone cannot ensure per-applicant object placement. F is wrong because SageMaker Clarify generates fairness and bias reports and does not satisfy document extraction, PII redaction, human review routing, or FM-driven loan assessment. Because the question asks for exactly three options, D must yield to A, B, and E, which each satisfy a named mandatory requirement rather than only a supporting storage concern.

Community Comment Notes

Gaurav955 posted 'ABD is final answer', and ansh0692 supported ABD by arguing that Glue Data Quality is batch-oriented rather than transactional; that critique of E is reasonable, but the chosen set still omits the Bedrock approval orchestration required by the scenario. AndreyShne24 focused on the decisive point that only B handles the PII requirement, which correctly pushes the answer toward ABE. QuinnGT similarly explained that A supplies managed extraction and human review, B redacts before the model call, and E provides durable orchestration and execution history. SankarKumar and de1612d favored ADE for residency and orchestration, but ADE drops B and therefore leaves the explicit pre-inference PII-redaction requirement unmet.

Official Reference

These links open the official vendor documentation in a new tab.

Exam Strategy

Match each hard requirement to a dedicated service and eliminate options that cover only storage or reporting. Since the prompt says choose three, reserve one slot for PII redaction (B), one for extraction plus in-Region human review (A), and one for orchestration and Bedrock approval (E); D is a plausible compliance add-on but not one of the three mandatory capabilities.

Frequently Asked Questions

Why is Amazon S3 object tagging in option D not enough for Regional data residency?

Tagging supports audit classification, but an S3 object remains in its bucket's Region; you still need Region-scoped buckets and IAM, plus the PII redaction and Bedrock workflow supplied by A, B, and E.

What role does AWS Glue Data Quality play in the ABE solution?

It validates structured fields extracted from each document before Step Functions builds the Bedrock prompt; low-confidence extraction pages are routed to humans by Textract and A2I in option A.

More AIP-C01 FAQ →

Discussion 9 contributions

Compiled from public community threads and reader submissions, lightly edited for clarity. Handles and vote tallies are shown as recorded, and comments reflect the views of individual contributors.

From public community threads 9

  1. AndreyShne24 👍 2
    Although D is a good answer as well to be ADE but only B is handling the PII requirement, so my vote goes to ABE
  2. jakie22332 👍 2
    this is the correct answer
  3. Gaurav955 👍 2
    ABD is final answer
  4. QuinnGT 👍 1
    A:supplies managed document extraction and routes low-confidence Textract results to people for review. B:performs PII redaction before the model call and adds output controls and Region-scoped authorization. E:validates the extracted structure and gives the multi-stage process a durable orchestration and execution history before invoking Bedrock. D is tempting because it mentions same-Region S3 and tags, but tags alone are not an audit trail, and B/E more directly fulfill mandatory processing requirements.
  5. de1612d 👍 1
    A: Use Amazon Textract with Amazon Augmented AI to extract structured data from scanned documents and route low-confidence results to human reviewers. D: Store documents in Amazon S3 with region-specific controls and object tagging to ensure data residency compliance and auditability. E: Use AWS Step Functions and AWS Glue Data Quality to orchestrate the workflow, validate extracted data, and invoke Amazon Bedrock for loan
Show 4 more comments
  1. ansh0692 👍 1
    AWS Glue Data Quality is designed for batch data lake transformations, not low-latency transactional document validation. Additionally, this option completely lacks a mechanism for routing low-confidence OCR extractions to human reviewers in-region. Hence, ABD
  2. GiorgioGss 👍 1
    C - Kendra/OpenSearch for semantic search, not structured field extraction E - Glue Data Quality validates data schemas, not loan decision workflows F - SageMaker Clarify for ML bias (not Bedrock FM outputs)
  3. SankarKumar 👍 1
    Scalable orchestration High availability (99.9%) Clean input → better FM output
  4. 67bdb19 👍 1 Answered B
    This is a tough one. My best educated guess is B.

Reader comments & corrections

No reader corrections yet — spotted an error or have a note? Share it below.

Log in to comment, report an error, or add a note about this question.

How Can AgentCore Query Cross-Account DynamoDB While Preserving User Identity?

S-grade

Which API Gateway setup streams Bedrock responses with least operational overhead?

S-grade

Resilient Retry and Streaming Strategies for Amazon Bedrock

S-grade

Bedrock Prompt Management & Guardrails Detection

S-grade

Practice All AIP-C01 Questions

Access 110 questions with complete answers and detailed explanations.

View Full AIP-C01 Practice Test →

← Back to AIP-C01 Study Guide