34 free AWS AIP-C01 practice questions from the CertStash pack of 136, each with the correct answer and a full explanation. The rule is the same for every exam: ten free on the product page, 34 free here, all 136 in the pack, and each step includes the one before it.
Get all 136 questions (US$39) · Download the first 10 as a PDF
Question 1
A retail company has a generative AI (GenAI) product recommendation application that uses Amazon Bedrock. The application suggests products to customers based on browsing history and demographics. The company needs to implement fairness evaluation across multiple demographic groups to detect and measure bias in recommendations between two prompt approaches. The company wants to collect and monitor fairness metrics in real time. The company must receive an alert if the fairness metrics show a discrepancy of more than 15% between demographic groups. The company must receive weekly reports that compare the performance of the two prompt approaches.
Which solution will meet these requirements with the LEAST custom development effort?
Show answer and explanation
Correct answer: B. Create the two prompt variants in Amazon Bedrock Prompt Management. Use Amazon Bedrock Flows to deploy the prompt variants with defined traffic allocation. Configure Amazon Bedrock guardrails that have content filters to monitor demographic fairness. Set up Amazon CloudWatch alarms on the GuardrailContentSource dimension that use InvocationsIntervened metrics to detect recommendation discrepancy threshold violations.
Option B directly addresses fairness evaluation with the least custom development by using Amazon Bedrock's built-in Prompt Management for variants, Bedrock Flows for traffic allocation, and Bedrock Guardrails specifically designed to monitor demographic fairness through content filters. The guardrails' InvocationsIntervened metrics on GuardrailContentSource dimensions can be configured with CloudWatch alarms for the 15% threshold violation detection. This approach minimizes custom development compared to building custom metrics and Lambda post-processing functions.
Why the other options are wrong
- A. Requires significant custom development building post-processing analysis functions and custom fairness metrics calculation with Lambda, increasing maintenance burden.
- C. SageMaker Clarify is designed for one-time model evaluation jobs, not real-time monitoring with alerts, and requires manual integration work rather than native Bedrock guardrail support.
- D. Model evaluation jobs are batch-oriented for comparing variants, not suited for real-time metric collection and alerting on fairness discrepancies.
Question 2
A finance company is developing an AI assistant to help clients plan investments and manage their portfolios. The company identifies several high-risk conversation patterns such as requests for specific stock recommendations or guaranteed returns. High-risk conversation patterns could lead to regulatory violations if the company cannot implement appropriate controls.
The company must ensure that the AI assistant does not provide inappropriate financial advice, generate content about competitors, or make claims that are not factually grounded in the company's approved financial guidance. The company wants to use Amazon Bedrock Guardrails to implement a solution.
Which combination of steps will meet these requirements? (Choose three.)
Show answer and explanation
Correct answer: A, D, F. Add the high-risk conversation patterns to a denied topics guardrail; Add the names of competitors as custom word filters. Set the input and output actions to block; Set a high grounding score threshold.
Option A addresses high-risk conversation patterns by adding them to denied topics guardrails. Option D blocks competitor names using custom word filters with input/output block actions. Option F sets a high grounding score threshold to ensure responses are factually grounded in approved financial guidance. Together, these three options comprehensively address preventing inappropriate financial advice, blocking competitor content, and ensuring factual accuracy without allowing unreliable content.
Why the other options are wrong
- B. Content filters are for filtering TYPES of content (violence, hate speech), not specific conversation patterns; denied topics guardrails are the correct mechanism.
- C. Custom word filters (option D) are more appropriate and direct for blocking competitor names than content filters designed for content categories.
- E. A low grounding score threshold would allow poorly grounded claims, directly opposing the requirement to ensure factually grounded guidance.
Question 3
A company has deployed an AI assistant as a React application that uses AWS Amplify, an AWS AppSync GraphQL API, and Amazon Bedrock Knowledge Bases. The application uses the GraphQL API to call the Amazon Bedrock RetrieveAndGenerate API for knowledge base interactions.
The company configures an AWS Lambda resolver to use the RequestResponse invocation type.
Application users report frequent timeouts and slow response times. Users report these problems more frequently for complex questions that require longer processing.
The company needs a solution to fix these performance issues and enhance the user experience.
Which solution will meet these requirements?
Show answer and explanation
Correct answer: A. Use AWS Amplify AI Kit to implement streaming responses from the GraphQL API and to optimize client-side rendering.
AWS Amplify AI Kit provides native support for streaming responses from GraphQL APIs and optimizes client-side rendering for long-running AI operations. This directly addresses the timeout and latency issues caused by complex questions requiring longer processing. The streaming approach allows responses to begin flowing to the client immediately rather than waiting for complete computation, significantly improving perceived performance and user experience without architectural changes.
Why the other options are wrong
- B. Simply increasing timeout values treats the symptom but doesn't solve the underlying slow response problem; exponential backoff on timeouts won't improve performance for intrinsically slow operations.
- C. SQS-based asynchronous polling adds significant architectural complexity and latency (polling overhead) compared to native streaming, which is inefficient for real-time user interactions.
- D. Switching APIs and adding WebSocket complexity is unnecessarily complicated when Amplify AI Kit provides streaming support natively integrated with existing GraphQL infrastructure.
Question 4
An ecommerce company operates a global product recommendation system that needs to switch between multiple foundation models (FM) in Amazon Bedrock based on regulations, cost optimization, and performance requirements. The company must apply custom controls based on proprietary business logic, including dynamic cost thresholds, AWS Region-specific compliance rules, and real-time A/B testing across multiple FMs. The system must be able to switch between FMs without deploying new code. The system must route user requests based on complex rules including user tier, transaction value, regulatory zone, and real-time cost metrics that change hourly and require immediate propagation across thousands of concurrent requests.
Which solution will meet these requirements?
Show answer and explanation
Correct answer: C. Configure an AWS Lambda function to fetch routing configurations from the AWS AppConfig Agent for each user request. Run business logic in the Lambda function to select the appropriate FM for each request. Expose the FM through a single Amazon API Gateway REST API endpoint.
Option C uses AWS AppConfig Agent to fetch routing configurations dynamically for each request, enabling immediate propagation of hourly cost threshold changes and regulatory updates across concurrent requests without code deployment. The Lambda function executes business logic to evaluate user tier, transaction value, regulatory zone, and real-time cost metrics against the dynamically fetched configuration. This architecture supports complex rule evaluation, real-time updates, and model switching without redeployment, meeting all requirements with minimal operational friction.
Why the other options are wrong
- A. Environment variables require Lambda redeployment to update routing rules; they don't support hourly changes or immediate propagation to thousands of concurrent requests.
- B. API Gateway request transformation templates cannot handle complex business logic like dynamic cost threshold evaluation and regulatory rule application; stage variables also require manual updates.
- D. Using Lambda authorizers for routing logic is an incorrect pattern; authorizers are designed for authentication/authorization, not request routing or business logic execution.
Question 5
A company is developing an internal generative AI (GenAI) assistant that uses Amazon Bedrock to summarize corporate documents for multiple business units. The GenAI assistant must generate responses in a consistent format that includes a document summary, classification of business risks, and terms that are flagged for review. The GenAI assistant must adapt the tone of responses for each user's business unit, such as legal, human resources, or finance. The GenAI assistant must block hate speech, inappropriate topics, and sensitive information such as personal health information.
The company needs a solution to centrally manage prompt variants across business units and teams. The company wants to minimize ongoing orchestration efforts and maintenance for post-processing logic. The company also wants to have the ability to adjust content moderation criteria for the GenAI assistant over time.
Which solution will meet these requirements with the LEAST maintenance overhead?
Show answer and explanation
Correct answer: A. Use Amazon Bedrock Prompt Management to configure reusable templates and business unit-specific prompt variants. Apply Amazon Bedrock guardrails that have category filters and sensitive term lists to block prohibited content.
Option A uses Amazon Bedrock Prompt Management to create reusable templates and business unit-specific variants, centralizing management across teams with minimal orchestration overhead. Amazon Bedrock Guardrails with category filters and sensitive term lists provide built-in content moderation without custom post-processing logic. This combination supports consistent formatting, tone adaptation per business unit, and adjustable content moderation criteria over time, all managed centrally with the least maintenance burden.
Why the other options are wrong
- B. Audience-based threshold tuning in guardrails and internal administration APIs add complexity without corresponding benefit over standard category filters.
- C. Instruction injection, DynamoDB storage, Step Functions validation, and Amazon Comprehend post-processing create significant custom maintenance overhead and orchestration complexity.
- D. Custom templates in DynamoDB, multiple Lambda functions, and post-processing with Comprehend introduce unnecessary operational overhead compared to native Bedrock Prompt Management and Guardrails.
Question 6
A financial services company is building a customer support application that retrieves relevant financial regulation documents from a database based on semantic similarities to user queries. The application must integrate with Amazon Bedrock to generate responses. The application must be able to search documents that are in English, Spanish, and Portuguese. The application must filter documents by metadata such as publication date, regulatory agency, and document type.
The database stores approximately 10 million document embeddings. To minimize operational overhead, the company wants a solution that minimizes management and maintenance effort. The application must provide low-latency responses for real-time customer interactions.
Which solution will meet these requirements?
Show answer and explanation
Correct answer: A. Use Amazon OpenSearch Serverless to provide vector search capabilities and metadata filtering. Connect to Amazon Bedrock Knowledge Bases to enable Retrieval Augmented Generation (RAG) capabilities that use an Anthropic Claude foundation model (FM).
Amazon OpenSearch Serverless provides vector search capabilities with built-in metadata filtering for publication date, regulatory agency, and document type across 10 million embeddings. It supports multi-language semantic search (English, Spanish, Portuguese) natively and integrates seamlessly with Amazon Bedrock Knowledge Bases for RAG capabilities using Anthropic Claude models. Serverless architecture minimizes operational overhead while providing low-latency retrieval for real-time customer interactions.
Why the other options are wrong
- B. Aurora PostgreSQL with pgvector requires manual database provisioning, scaling, and operational management, increasing overhead compared to serverless offerings.
- C. S3 Vectors does not support metadata filtering, a key requirement for filtering by publication date, regulatory agency, and document type.
- D. Neptune Analytics is a graph database designed for relationship queries, not optimized for vector similarity search or the operational simplicity required for this use case.
Question 7
A medical company is building a generative AI (GenAI) application that uses RAG to provide evidence-based medical information. The application uses Amazon OpenSearch Service to retrieve vector embeddings. Users report that searches frequently miss results that contain exact medical terms and acronyms and return too many semantically similar but irrelevant documents. The company needs to improve retrieval quality and maintain low end user latency, even as the document collection grows to millions of documents.
Which solution will meet these requirements with the LEAST operational overhead?
Show answer and explanation
Correct answer: A. Configure hybrid search by combining vector similarity with keyword matching to improve semantic understanding and exact term and acronym matching.
Hybrid search combines vector similarity with keyword matching, allowing exact matches on medical terms and acronyms while maintaining semantic understanding. This approach directly addresses both problems: missing results containing exact medical terminology and excessive semantically similar irrelevant documents. Implementing hybrid search in Amazon OpenSearch Service requires minimal operational overhead (configuration change) while maintaining low latency and supporting growth to millions of documents.
Why the other options are wrong
- B. Increasing vector dimensions and post-processing filtering doesn't address the underlying issue that pure semantic search misses exact term matches and requires Lambda post-processing overhead.
- C. Replacing OpenSearch with Amazon Kendra introduces significant migration effort and operational changes; Kendra is designed for enterprise document search, not specialized RAG tasks.
- D. Two-stage retrieval with SageMaker ML re-ranking adds operational complexity, management overhead, and additional latency, contradicting the low-overhead requirement.
Question 8
A company runs a generative AI (GenAI)-powered summarization application in an application AWS account that uses Amazon Bedrock. The application architecture includes an Amazon API Gateway REST API that forwards requests to AWS Lambda functions that are attached to private VPC subnets. The application summarizes sensitive customer records that the company stores in a governed data lake in a centralized data storage account. The company has enabled Amazon S3, Amazon Athena, and AWS Glue in the data storage account.
The company must ensure that calls that the application makes to Amazon Bedrock use only private connectivity between the company's application VPC and Amazon Bedrock. The company's data lake must provide fine-grained column-level access across the company's AWS accounts.
Which solution will meet these requirements?
Show answer and explanation
Correct answer: A. In the application account, create interface VPC endpoints for Amazon Bedrock runtimes. Run Lambda functions in private subnets. Use IAM conditions on inference and data-plane policies to allow calls only to approved endpoints and roles. In the data storage account, use AWS Lake Formation LF-tag-based access control to create table and column-level cross-account grants.
Option A creates interface VPC endpoints for Amazon Bedrock runtime in the application account, ensuring private connectivity without internet exposure. Lambda functions in private subnets use these endpoints exclusively with IAM conditions restricting access to approved endpoints and roles. AWS Lake Formation LF-tag-based access control provides fine-grained column-level cross-account access to the data lake, addressing both the private connectivity requirement for Bedrock and the governed data lake access requirement.
Why the other options are wrong
- B. NAT gateways route traffic through public internet, violating the private connectivity requirement; S3 bucket policies and ACLs lack fine-grained column-level control.
- C. Invoking Bedrock through public endpoints violates the private connectivity requirement; database-level grants lack column-level granularity required for fine-grained access.
- D. Public fallback for cross-Region reads bypasses the private connectivity requirement; IAM path-based policies alone cannot enforce column-level access control like Lake Formation provides.
Question 9
A media company must use Amazon Bedrock to implement a robust governance process for AI-generated content. The company needs to manage hundreds of prompt templates. Multiple teams use the templates across multiple AWS Regions to generate content. The solution must provide version control with approval workflows that include notifications for pending reviews. The solution must also provide detailed audit trails that document prompt activities and consistent prompt parameterization to enforce quality standards.
Which solution will meet these requirements?
Show answer and explanation
Correct answer: B. Use Amazon Bedrock Prompt Management to implement version control. Configure AWS CloudTrail for audit logging. Use IAM policies to control approval permissions. Create parameterized prompt templates by specifying variables.
Amazon Bedrock Prompt Management is the native service designed specifically for managing prompt templates with version control and approval workflows. CloudTrail provides comprehensive audit logging of all prompt activities. IAM policies control who can approve prompts, and parameterized templates with variables enforce consistent quality standards across teams and regions. This is the purpose-built solution for the company's governance requirements.
Why the other options are wrong
- A. Bedrock Studio is for interactive development, not enterprise prompt management; CloudWatch metrics don't provide approval workflows; DynamoDB with Lambda is a custom solution lacking built-in version control.
- C. Step Functions and S3 lack native prompt versioning and approval features; tags are not version control; this requires custom development.
- D. SageMaker Canvas is for different use cases; CloudFormation manages infrastructure not prompts; AWS Config doesn't enforce approval workflows.
Question 10
A company is developing a customer support application that uses Amazon Bedrock foundation models (FMs) to provide real-time AI assistance to the company's employees. The application must display AI-generated responses character by character as the responses are generated. The application needs to support thousands of concurrent users with minimal latency. The responses typically take 15 to 45 seconds to finish.
Which solution will meet these requirements?
Show answer and explanation
Correct answer: A. Configure an Amazon API Gateway WebSocket API with an AWS Lambda integration. Configure the WebSocket API to invoke the Amazon Bedrock InvokeModelWithResponseStream API and stream partial responses through WebSocket connections.
WebSocket APIs enable persistent bidirectional connections ideal for streaming character-by-character output. The InvokeModelWithResponseStream API is specifically designed for streaming partial model outputs. Lambda can handle thousands of concurrent WebSocket connections efficiently. This architecture supports real-time display of AI responses with minimal latency and scales to thousands of concurrent users.
Why the other options are wrong
- B. REST APIs with polling every 100ms creates unnecessary latency and load; standard InvokeModel is not designed for streaming; polling is inefficient for long 15-45 second responses.
- C. Direct client connections to Bedrock expose AWS credentials and lack proper API management; this violates security best practices and doesn't scale well.
- D. HTTP API with caching and pagination defeats the purpose of real-time streaming; DynamoDB persistence adds latency; paginated GET requests don't provide character-by-character streaming experience.
What people say about the full AIP-C01 pack 5.0 from 1
AIP-C01 passed! Happy with the result.
Question 11
A healthcare company is using Amazon Bedrock to develop a real-time patient care AI assistant to respond to queries for separate departments that handle clinical inquiries, insurance verification, appointment scheduling, and insurance claims. The company wants to use a multi-agent architecture.
The company must ensure that the AI assistant is scalable and can onboard new features for patients. The AI assistant must be able to handle thousands of parallel patient interactions. The company must ensure that patients receive appropriate domain-specific responses to queries.
Which solution will meet these requirements?
Show answer and explanation
Correct answer: A. Isolate data for each agent by using separate knowledge bases. Use IAM filtering to control access to each knowledge base. Deploy a supervisor agent to perform natural language intent classification on patient inquiries. Configure the supervisor agent to route queries to specialized collaborator agents to respond to department-specific queries. Configure each specialized collaborator agent to use Retrieval Augmented Generation (RAG) with the agent's department-specific knowledge base.
A supervisor agent performing natural language intent classification routes to specialized collaborator agents, each with department-specific knowledge bases and IAM-controlled access. This architecture is scalable (supervisor handles routing, collaborators handle domain logic), enables feature onboarding (add new collaborator agents), supports thousands of parallel interactions (Step Functions/Lambda scale automatically), and ensures domain-specific responses (each agent has isolated, relevant knowledge). This hierarchical design separates concerns optimally.
Why the other options are wrong
- B. Multiple supervisor agents create redundant routing logic and manual handoff processes between them introduce bottlenecks and increased complexity.
- C. A single general-purpose agent with multiple action groups cannot scale intent classification or maintain clean separation of concerns; rule-based routing is brittle and inflexible.
- D. Parallel independent supervisors with merged responses creates conflicting recommendations and is operationally inefficient; a shared knowledge base violates domain isolation and causes irrelevant results.
Question 12
A company is building a generative AI (GenAI) application that produces content based on a variety of internal and external data sources. The company wants to ensure that the generated output is fully traceable. The application must support data source registration and enable metadata tagging to attribute content to its original source. The application must also maintain audit logs of data access and usage throughout the pipeline.
Which solution will meet these requirements?
Show answer and explanation
Correct answer: D. Use AWS Glue Data Catalog to register all data sources. Apply metadata tags to attribute data sources. Use AWS CloudTrail to log access and activity across services.
AWS Glue Data Catalog is the primary service for registering and managing data sources with comprehensive metadata tagging capabilities, enabling source attribution throughout the data pipeline. CloudTrail provides the complete audit trail of access and activity across all services, creating a comprehensive lineage record from source through generation. This combination provides full traceability with metadata tags directly linked to registered sources and complete audit logging across the entire pipeline, meeting all traceability requirements.
Why the other options are wrong
- A. Lake Formation is for access control rather than data lineage; lacks built-in attribution and tagging capabilities for source identification.
- B. CloudWatch Logs monitors behavior but does not provide data source registration or comprehensive audit trails; lacks metadata tagging for attribution.
- C. S3 object tagging is insufficient for comprehensive data lineage; lacks centralized source registration and does not provide the complete audit trail that CloudTrail offers.
Question 13
A company is using Amazon Bedrock to build a customer-facing AI assistant to handle sensitive customer inquiries. The company must use defense-in-depth safety controls to block sophisticated prompt injection attacks. The company must keep audit logs of all safety interventions.
The AI assistant must have cross-Region failover capabilities.
Which solution will meet these requirements?
Show answer and explanation
Correct answer: A. Configure Amazon Bedrock guardrails to use content filters to protect against prompt injection attacks. Set the content filters to high. Use a guardrail profile to implement cross-Region guardrail inference. Use Amazon CloudWatch Logs with custom metrics to capture detailed guardrail intervention events.
Amazon Bedrock guardrails with high-level content filters provide defense-in-depth protection against prompt injection attacks with sophisticated filtering logic. Guardrail profiles enable cross-Region inference, providing the required failover capability across regions. CloudWatch Logs with custom metrics capture detailed guardrail intervention events for audit logging. This approach combines specialized guardrail technology with proper audit trails and cross-Region failover in a cohesive solution.
Why the other options are wrong
- B. AWS WAF operates at network layer and cannot detect sophisticated prompt injection within model inputs; CloudTrail logs API calls but not detailed guardrail interventions.
- C. Amazon Comprehend custom classification requires custom training and maintenance; API Gateway validation adds latency; does not provide cross-Region failover.
- D. Word filters for attack patterns are brittle and easily bypassed; custom content filters require ongoing maintenance; guardrail replication across regions is not a standard feature.
Question 14
A company is using Amazon Bedrock to design an application to help researchers apply for grants. The application is based on an Amazon Nova Pro foundation model (FM). The application contains four required inputs and must provide responses in a consistent text format. The company wants to receive a notification in Amazon Bedrock if a response contains bullying language. However, the company does not want to block all flagged responses.
The company creates an Amazon Bedrock flow that takes an input prompt and sends it to the Amazon Nova Pro FM. The Amazon Nova Pro FM provides a response.
Which additional steps must the company take to meet these requirements? (Choose two.)
Show answer and explanation
Correct answer: A, D. Use Amazon Bedrock Prompt Management to specify the required inputs as variables. Select an Amazon Nova Pro FM. Specify the output format for the response. Add the prompt to the prompts node of the flow; Create an Amazon Bedrock guardrail that applies the insults content filter. Set the filter response to detect. Add the guardrail to the prompts node of the flow.
Option A uses Bedrock Prompt Management to define required inputs as variables and specify output format, which ensures consistent text formatting. Option D creates a guardrail with the insults filter set to detect mode, which sends notifications about bullying language without blocking responses, meeting the requirement to detect but not block flagged content. Together these ensure structured prompts and content detection without blocking.
Why the other options are wrong
- B. Setting the filter response to block contradicts the requirement to not block flagged responses; the requirement specifically states company doesn't want to block all flagged responses.
- C. Prompt routers are for directing to different models, not for defining inputs and output formats; this doesn't address the consistent format requirement.
- E. Application inference profiles specify model configurations but don't define required inputs as variables or output format constraints; this is not the appropriate mechanism for prompt structure.
Question 15
A financial technology company is using Amazon Bedrock to build an assessment system for the company's customer service AI assistant. The AI assistant must provide financial recommendations that are factually accurate, compliant with financial regulations, and conversationally appropriate. The company needs to combine automated quality evaluations at scale with targeted human reviews of critical interactions.
What solution will meet these requirements?
Show answer and explanation
Correct answer: B. Configure Amazon Bedrock evaluations that use Anthropic Claude Sonnet as a judge model to assess response accuracy and appropriateness. Configure custom Amazon Bedrock guardrails to check responses for compliance with financial policies. Add Amazon Augmented AI (Amazon A2I) human reviews for flagged critical interactions.
Amazon Bedrock evaluations with Claude Sonnet as a judge model automates large-scale quality assessment for accuracy and conversational appropriateness. Custom guardrails provide automated compliance checking against financial policies. Amazon A2I human reviews flag critical interactions that fail automated checks, combining efficient automated evaluation with targeted human oversight. This hybrid approach meets all requirements: accuracy, compliance, conversational quality, and scalability without relying entirely on manual review.
Why the other options are wrong
- A. Manual scoring by all responses is not scalable and cannot handle the volume needed for a production assessment system.
- C. Amazon Lex is a conversational UI tool, not an evaluation framework; static compliance databases lack the flexibility of guardrails, and end-user reviews are not systematic quality assessment.
- D. CloudWatch pattern monitoring and manual human review lack the automated evaluation capabilities and cannot assess accuracy or conversational appropriateness systematically.
Question 16
A company is building an AI advisory application by using Amazon Bedrock. The application will provide recommendations to customers. The company needs the application to explain its reasoning process and cite specific sources for data. The application must retrieve information from company data sources and show step-by-step reasoning for recommendations. The application must also link data claims to source documents and maintain response latency under 3 seconds.
Which solution will meet these requirements with the LEAST operational overhead?
Show answer and explanation
Correct answer: A. Use Amazon Bedrock Knowledge Bases with source attribution enabled. Use the Anthropic Claude Messages API with RAG to set highrelevance thresholds for source documents. Store reasoning and citations in Amazon S3 for auditing purposes.
Amazon Bedrock Knowledge Bases with source attribution enabled provides built-in mechanisms to cite specific sources and explain reasoning within responses. The integration with Anthropic Claude's RAG capabilities automatically handles retrieval and citation with relevance thresholds, maintaining sub-3-second latency while offering least operational overhead. Source attribution is a native Knowledge Bases feature requiring no custom implementation, and S3 storage for auditing is straightforward and scalable.
Why the other options are wrong
- B. Extended thinking with 4,000-token budget increases latency and token costs; requires manual citation management and storage in DynamoDB.
- C. Custom SageMaker configuration introduces operational complexity; AWS Lambda processing adds latency; RDS sourcing for citations is indirect.
- D. Chain-of-thought reasoning does not inherently provide source citations; custom retrieval tracking requires additional implementation beyond standard Knowledge Bases features.
Question 17
A company is developing a generative AI (GenAI)-powered customer support application that uses Amazon Bedrock foundation models (FMs).
The application must maintain conversational context across multiple interactions with the same user. The application must run clarification workflows to handle ambiguous user queries. The company must store encrypted records of each user conversation to use for personalization.
The application must be able to handle thousands of concurrent users while responding to each user quickly.
Which solution will meet these requirements?
Show answer and explanation
Correct answer: B. Use an AWS Step Functions Standard workflow to orchestrate clarification workflows. Include Wait for a Callback patterns to manage the workflows. Store conversation history in Amazon DynamoDPurchase on-demand capacity and configure server-side encryption.
Step Functions Standard workflows with Wait for a Callback patterns are designed for orchestrating multi-step conversations with clarification logic. DynamoDB with on-demand capacity handles thousands of concurrent users without provisioning overhead, and server-side encryption meets compliance requirements. This architecture maintains conversational context per user while managing complex clarification workflows at scale.
Why the other options are wrong
- A. Express workflows lack the callback and wait mechanics needed for interactive clarification workflows; RDS with session IDs is less suited for high-concurrency, low-latency conversation tracking.
- C. Storing individual JSON files in S3 introduces high latency for conversation retrieval and is not appropriate for maintaining state across thousands of concurrent users.
- D. SQS is asynchronous and not ideal for maintaining real-time conversational context; ElastiCache lacks the durability guarantees needed for encrypted conversation records.
Question 18
A financial services company uses an AI application to process financial documents by using Amazon Bedrock. During business hours, the application handles approximately 10,000 requests each hour, which requires consistent throughput.
The company uses the CreateProvisionedModelThroughput API to purchase provisioned throughput. Amazon CloudWatch metrics show that the provisioned capacity is unused while on-demand requests are being throttled. The company finds the following code in the application: python response = bedrock_runtime.invoke_model(modelId="anthropic.claude-v2", body=json.dumps(payload))
The company needs the application to use the provisioned throughput and to resolve the throttling issues.
Which solution will meet these requirements?
Show answer and explanation
Correct answer: B. Replace the model ID parameter with the ARN of the provisioned model that the CreateProvisionedModelThroughput API returns.
The application code uses the string model ID 'anthropic.claude-v2' instead of the ARN returned by CreateProvisionedModelThroughput API, causing Bedrock to route requests to on-demand capacity instead of provisioned throughput. Replacing the model ID with the provisioned model ARN directs all requests to the allocated provisioned capacity, resolving throttling and utilizing the purchased throughput. This is the specific configuration required by Bedrock's API.
Why the other options are wrong
- A. Increasing MUs does not resolve the issue if requests are not being routed to provisioned capacity; the problem is routing, not capacity.
- C. Exponential backoff handles throttling exceptions but does not redirect traffic to provisioned throughput; merely masks the underlying routing issue.
- D. InvokeModelWithResponseStream is for streaming responses but does not affect which capacity tier (provisioned vs on-demand) handles requests.
Question 19
A financial services company uses multiple foundation models (FMs) through Amazon Bedrock for its generative AI (GenAI) applications. To comply with a new regulation for GenAI use with sensitive financial data, the company needs a token management solution.
The token management solution must proactively alert when applications approach model-specific token limits. The solution must also process more than 5,000 requests each minute and maintain token usage metrics to allocate costs across business units.
Which solution will meet these requirements?
Show answer and explanation
Correct answer: A. Develop model-specific tokenizers in an AWS Lambda function. Configure the Lambda function to estimate token usage before sending requests to Amazon Bedrock. Configure the Lambda function to publish metrics to Amazon CloudWatch and trigger alarms when requests approach thresholds. Store detailed token usage in Amazon DynamoDB to report costs.
Option A directly addresses all requirements: Lambda functions with model-specific tokenizers can estimate token usage before sending requests to Bedrock, CloudWatch metrics and alarms handle proactive alerting for approaching token limits, and DynamoDB stores detailed usage data for cost allocation. Lambda scales efficiently to handle 5,000+ requests per minute. This approach provides complete visibility and control over token consumption for compliance with financial regulations.
Why the other options are wrong
- B. Bedrock Guardrails with token quota policies would reject requests reactively rather than proactively alert when approaching limits, and EventBridge rules lack the granular token estimation capability needed.
- C. Using SQS dead-letter queues and analyzing errors post-failure is reactive rather than proactive, and CloudWatch Logs Insights cannot provide accurate token estimation before requests are sent.
- D. API Gateway throttling is coarse-grained and cannot perform model-specific token estimation, nor can it effectively track detailed token usage metrics for cost allocation across business units.
Question 20
A company uses an AI assistant application to summarize the company's website content and provide information to customers. The company plans to use Amazon Bedrock to give the application access to a foundation model (FM).
The company needs to deploy the AI assistant application to a development environment and a production environment. The solution must integrate the environments with the FM. The company wants to test the effectiveness of various FMs in each environment. The solution must provide product owners with the ability to easily switch between FMs for testing purposes in each environment.
Which solution will meet these requirements?
Show answer and explanation
Correct answer: C. Create one AWS CDK application. Configure the application to invoke the Amazon Bedrock FMs by using the aws_bedrock.FoundationModel.fromFoundationModelId() method. Create a pipeline in AWS CodePipeline pipeline that has a deployment stage for each environment that uses AWS CodeBuild deploy actions.
Option C creates a single AWS CDK application that uses the FoundationModel.fromFoundationModelId() method, allowing easy switching between foundation models by changing the model ID without code redeployment. A single CodePipeline with deployment stages for each environment enables consistent infrastructure while allowing environment-specific configuration through deployment parameters. This approach provides maximum flexibility for testing different FMs in dev and prod while maintaining a single source of truth for infrastructure as code.
Why the other options are wrong
- A. Using ProvisionedModel.fromProvisionedModelArn() requires pre-provisioned capacity and is unnecessarily complex for testing multiple FMs; multiple pipelines add operational overhead.
- B. Creating separate CDK applications for each environment violates DRY principles, doubles maintenance burden, and makes it harder to ensure consistency between dev and prod.
- D. Manually recreating resources for the development environment eliminates infrastructure-as-code benefits and creates drift between environments, defeating the purpose of CDK.
Question 21
A financial services company is developing a generative AI (GenAI) application that serves both premium customers and standard customers. The application uses AWS Lambda functions behind an Amazon API Gateway REST API to process requests. The company needs to dynamically switch between AI models based on which customer tier each user belongs to. The company also wants to perform A/B testing for new features without redeploying code. The company needs to validate model parameters like temperature and maximum token limits before applying changes.
Which solution will meet these requirements with the LEAST operational overhead?
Show answer and explanation
Correct answer: C. Use AWS AppConfig to manage model configurations. Use feature flags to perform A/B testing. Define JSON schema validation rules for model parameters. Configure Lambda functions to retrieve configurations by using the AWS AppConfig Agent.
AWS AppConfig is purpose-built for managing application configurations with built-in support for feature flags, JSON schema validation, and safe deployment strategies. It provides the least operational overhead by eliminating the need for custom polling logic, manual redeployments, or complex caching strategies. The AppConfig Agent integrates seamlessly with Lambda functions and handles validation rules for model parameters like temperature and token limits. Feature flags enable A/B testing without code changes, and deployment strategies provide safe rollout mechanisms.
Why the other options are wrong
- A. Parameter Store requires custom polling logic and EventBridge-triggered redeployments, adding operational complexity compared to AppConfig's native support.
- B. DynamoDB requires custom access pattern optimization and adds query latency on every request compared to AppConfig's optimized delivery model.
- D. ElastiCache with custom validation logic and manual TTL management introduces unnecessary complexity and operational overhead versus AppConfig's managed features.
Question 22
Example Corp provides a personalized video generation service that millions of enterprise customers use. Customers generate marketing videos by submitting prompts to the company's proprietary generative AI (GenAI) model. To improve output relevance and personalization, Example Corp wants to enhance the prompts by using customer-specific context such as product preferences, customer attributes, and business history.
The customers have strict data governance requirements. The customers must retain full ownership and control over their own data. The customers do not require real-time access. However, semantic accuracy must be high and retrieval latency must remain low to support customer experience use cases.
Example Corp wants to minimize architectural complexity in its integration pattern. Example Corp does not want to deploy and manage services in each customer's environment unless necessary.
Which solution will meet these requirements?
Show answer and explanation
Correct answer: C. Ensure that each customer configures an Amazon Bedrock knowledge base. Allow cross-account querying so Example Corp can retrieve structured data for prompt augmentation.
Amazon Bedrock knowledge bases support cross-account querying, enabling each customer to maintain full ownership and control of their data within their own AWS account while allowing Example Corp to retrieve augmentation data through secure cross-account access. This maintains customer data sovereignty, requires no service deployment in customer environments, supports low retrieval latency for semantic retrieval, and minimizes architectural complexity compared to federated or real-time alternatives. Knowledge bases provide the semantic search capability needed while respecting the data governance requirement.
Why the other options are wrong
- A. Amazon Q Business is designed for enterprise search and uses a different index model not optimized for GenAI prompt augmentation; the requirement for data accessor designation is less straightforward than cross-account Bedrock access.
- B. Deploying MCP servers for each customer violates the requirement to avoid deploying services in each customer's environment and adds significant operational complexity.
- D. Amazon Kendra is a search service, not optimized for GenAI semantic retrieval; crawling customer data sources creates data governance concerns; sharing Kendra indexes across accounts is not the intended design pattern.
Question 23
A company has a recommendation system. The system's applications run on Amazon EC2 instances. The applications make API calls to Amazon Bedrock foundation models (FMs) to analyze customer behavior and generate personalized product recommendations.
The system is experiencing intermittent issues. Some recommendations do not match customer preferences. The company needs an observability solution to monitor operational metrics and detect patterns of operational performance degradation compared to established baselines. The solution must also generate alerts with correlation data within 10 minutes when FM behavior deviates from expected patterns.
Which solution will meet these requirements?
Show answer and explanation
Correct answer: C. Enable Amazon CloudWatch Application Insights for the application resources. Create custom metrics for recommendation quality, token usage, and response latency by using the CloudWatch embedded metric format with dimensions for request types and user segments. Configure CloudWatch anomaly detection on the model metrics. Establish log pattern analysis by using CloudWatch Logs Insights.
Option C provides comprehensive observability specifically designed for application resources with Application Insights, creates custom metrics for recommendation quality and token usage, uses CloudWatch anomaly detection to automatically compare against baselines, and CloudWatch Logs Insights for pattern analysis, all alert within 10 minutes. This integrated approach gives operational performance visibility and detects FM behavior deviations within the required timeframe while remaining focused on the application's specific needs.
Why the other options are wrong
- A. Container Insights focuses on infrastructure metrics, not FM behavior; basic CloudWatch alarms lack anomaly detection and won't automatically detect deviations from baselines.
- B. X-Ray and CloudTrail provide tracing but not anomaly detection; QuickSight requires manual analysis and doesn't meet the 10-minute alert requirement for automated deviation detection.
- D. OpenSearch with Kinesis ingestion adds significant operational overhead and complexity; PPL queries require manual crafting rather than automatic anomaly detection.
Question 24
A healthcare company is developing a document management system that stores medical research papers in an Amazon S3 bucket. The company needs to build a comprehensive metadata framework that will improve search precision for a generative AI (GenAI) application that analyzes the research papers. The metadata framework must include document timestamps, author information, and research domain classifications.
The solution must maintain a consistent metadata structure across all uploaded documents. The solution must give foundation models (FMs) the ability to understand document context without accessing the full content.
Which solution will meet these requirements?
Show answer and explanation
Correct answer: A. Store document timestamps in Amazon S3 system metadata. Use S3 object tags to implement domain classification. Implement custom user-defined metadata to store author information.
Option A uses the appropriate S3 metadata mechanisms: system metadata (timestamps) is automatically managed by S3, S3 object tags provide a structured tagging system ideal for domain classification that is queryable and indexable, and custom user-defined metadata stores flexible author information. This combination creates a consistent, queryable metadata structure that foundation models can use to understand document context through metadata retrieval without accessing full document content, while maintaining simplicity and native S3 integration.
Why the other options are wrong
- B. Object Lock is designed for compliance retention, not metadata storage; access points are networking/permission constructs unrelated to metadata; this option misuses S3 services for their intended purposes.
- C. Inventory reports are for operational reporting, not metadata storage; access points don't implement classification; Storage Lens dashboards provide metrics, not metadata storage, this conflates operational tools with data management.
- D. Object Lock retention periods track legal holds, not document timestamps; Event Notifications trigger processes but don't store classification data, this option incorrectly applies compliance and event streaming features to metadata storage.
Question 25
A company is developing a generative AI (GenAI) application that analyzes customer service calls in real-time and generates suggested responses for human customer service agents. The application must process 500,000 concurrent calls during peak hours with less than 200 ms end-to-end latency for each suggestion. The company uses existing architecture to transcribe customer call audio streams. The application must not exceed a pre-defined monthly compute budget and must maintain auto scaling capabilities.
Which solution will meet these requirements?
Show answer and explanation
Correct answer: B. Deploy a low-latency, real-time optimized model on Amazon Bedrock. Purchase provisioned throughput and set up automatic scaling policies.
The requirement for 500,000 concurrent calls with sub-200ms latency demands a low-latency, real-time optimized model rather than a complex reasoning model. Amazon Bedrock with provisioned throughput provides predictable performance and cost control within the budget constraints. Automatic scaling policies enable handling peak loads while maintaining latency requirements. This setup avoids batch processing delays and SageMaker instance cold-start issues that would violate the latency SLA.
Why the other options are wrong
- A. Batch processing is incompatible with real-time suggestion requirements and large complex models add inference latency.
- C. SageMaker GPU endpoints require manual scaling management and lack the built-in auto-scaling efficiency of Bedrock provisioned throughput.
- D. Serverless endpoints and batch optimization directly conflict with real-time latency and concurrent call requirements.
Question 26
A company uses AWS Lambda functions to build an AI agent solution. A GenAI developer must set up a Model Context Protocol (MCP) server that accesses user information. The GenAI developer must also configure the AI agent to use the new MCP server. The GenAI developer must ensure that only authorized users can access the MCP server.
Which solution will meet these requirements?
Show answer and explanation
Correct answer: C. Use a Lambda function to host the MCP server. Create an Amazon API Gateway HTTP API that proxies requests to the Lambda function. Configure the AI agent solution to use the Streamable HTTP transport to make requests through the HTTP API. Use Amazon Cognito to enforce OAuth 2.1.
Hosting the MCP server in Lambda with API Gateway HTTP API proxy provides proper authorization through Amazon Cognito OAuth 2.1, ensuring only authorized users access the server. The Streamable HTTP transport is the appropriate mechanism for Lambda-to-API Gateway communication. This setup separates concerns, scales independently, and implements industry-standard OAuth 2.1 authorization, meeting all three requirements: MCP server access, agent configuration, and authorization.
Why the other options are wrong
- A. Asynchronous invocation is inappropriate for MCP which requires synchronous request-response communication patterns.
- B. STDIO transport requires the MCP server to run in the same process as the client, making separate Lambda deployment impossible.
- D. Lambda layers cannot host an MCP server independently; they are code dependencies, not separate services. This architecture violates the isolation and authorization requirements.
Question 27
A software company is using Amazon Q Business to build an AI assistant that allows employees to access company information and personal information by using natural language prompts. The company stores this information in an Amazon S3 bucket.
Each department in the company has a dedicated prefix in the S3 bucket. Each object name includes the S3 prefix of the department that it belongs to. Each department can belong to only a single group in AWS IAM Identity Center. Each employee belongs to a single department.
The company configures Amazon Q Business to access data stored in an S3 bucket as a data source. The company needs to ensure that the AI assistant respects access controls based on the user's IAM Identity Center group membership.
Which solution will meet this requirement with the LEAST operational overhead?
Show answer and explanation
Correct answer: B. Create a single JSON file named acl.json at the top level of the S3 bucket. Add access control entries that map each department's S3 prefix to its corresponding IAM Identity Center group. Indicate the location of the JSON file in the Access Control section of the data source settings.
Amazon Q Business natively supports ACL files for access control based on IAM Identity Center group membership. A single centralized acl.json file at the bucket root level with mappings between department S3 prefixes and IAM Identity Center groups provides the least operational overhead. This single point of control avoids maintaining multiple ACL files per department folder and integrates directly with Amazon Q's Access Control section in data source settings. No custom Lambda logic or complex IAM permission sets are required.
Why the other options are wrong
- A. Multiple acl.json files across department folders create maintenance overhead and scaling issues as departments grow.
- C. IAM permissions sets with StringNotEquals conditions are overly complex for this use case and don't integrate with Amazon Q's native ACL mechanism.
- D. metadata.json with AccessControlList objects is not a recognized Amazon Q access control pattern; the standard approach uses acl.json files.
Question 28
A legal research company has a Retrieval Augmented Generation (RAG) application that uses Amazon Bedrock and Amazon OpenSearch Service.
The application stores 768-dimensional vector embeddings for 15 million legal documents, including statutes, court rulings, and case summaries.
The company's current chunking strategy segments text into fixed-length blocks of 500 tokens. The current chunking strategy often splits contextually linked information such as legal arguments, court opinions, or statute references across separate chunks. Researchers report that generated outputs frequently omit key context or cite outdated legal information.
Recent application logs show a 40% increase in response times. The p95 latency metric exceeds 2 seconds. The company expects storage needs for the application to grow from 90 GB to 360 GB within a year.
The company needs a solution to improve retrieval relevance and system performance at scale.
Which solution will meet these requirements?
Show answer and explanation
Correct answer: C. Update the chunking strategy to use semantic boundaries such as complete legal arguments, clauses, or sections rather than fixed token limits. Regenerate vector embeddings to align with the new chunk structure.
The core issue is that fixed-length chunking splits semantically related legal content across boundaries, causing context loss and citation errors. Semantic chunking at legal boundaries (arguments, clauses, sections) preserves contextual relationships, improving retrieval relevance and reducing the need to search through fragmented information. Regenerating embeddings ensures vectors accurately represent the new semantic chunks, directly addressing both relevance and performance issues while scaling efficiently.
Why the other options are wrong
- A. Increasing vector dimensionality from 768 to 4,096 does not address the root cause of split contexts and will increase storage and latency without solving retrieval quality.
- B. Static pre-written summaries eliminate the dynamic retrieval capability needed for a RAG system and don't address the chunking problem causing information gaps.
- D. DynamoDB is not designed for vector similarity searches; keyword indexes lack the semantic understanding needed for complex legal document retrieval.
Question 29
A retail company is developing a customer service application that must process 10,000 daily queries about products, orders, and warranties. The application must be able to respond to queries about 50,000 product documents that are updated every day. The application must integrate with an order management API to check the status of orders and to help process returns. The application must maintain context throughout multi-turn interactions with customers. The company must collect complete audit trails for application responses.
Which solution will meet these requirements with the LEAST operational overhead?
Show answer and explanation
Correct answer: D. Use an Amazon Bedrock agent with action groups to integrate with the order management API. Associate an Amazon Bedrock knowledge base with the agent to search product documentation by using Retrieval Augmentation Generation (RAG). Enable trace events to capture audit trails.
Option D uses Amazon Bedrock agents with action groups for order management integration and a knowledge base with RAG for product document retrieval, which minimizes operational overhead. Agents automatically handle multi-turn context management, and trace events provide built-in audit trail collection without custom implementation. Knowledge bases handle daily updates of 50,000 documents efficiently through semantic search, eliminating the need to manage multiple fine-tuned models or custom orchestration logic.
Why the other options are wrong
- A. Fine-tuning multiple models per product category creates significant operational overhead and doesn't scale efficiently for 50,000 daily document updates.
- B. Custom continued pre-training is operationally intensive and Amazon Kendra requires manual orchestration with Step Functions to coordinate all components.
- C. SageMaker with containers, Kendra, and Step Functions introduces more operational complexity than using managed Bedrock agents and knowledge bases.
Question 30
An ecommerce company is using Amazon Bedrock to build a customer service AI assistant. The AI assistant needs to process over 50,000 customer inquiries every day. The AI assistant occasionally experiences traffic spikes of up to 150,000 inquiries every day during promotional events. Analysis shows that 40% of inquiries follow similar patterns that share the same context.
A GenAI developer must design a solution that will ensure low latency and consistent performance for the AI assistant during traffic spikes.
Which solution will meet these requirements MOST cost-effectively?
Show answer and explanation
Correct answer: A. Configure latency-optimized inference by setting the latency parameter to optimized in the performance configuration of the request to Amazon Bedrock. Use prompt caching to handle the repetitive inquiries.
Prompt caching in Amazon Bedrock directly addresses the 40% of inquiries with similar patterns that share context, reducing latency and costs by caching repeated input tokens. The latency-optimized inference parameter ensures consistent performance during spikes. This combination handles the core problem (repetitive contexts) cost-effectively without the overhead of provisioned throughput or cross-region complexity.
Why the other options are wrong
- B. Purchasing provisioned throughput for peak loads (150k) is expensive and unnecessary when 40% of traffic is repetitive and can be cached efficiently.
- C. Amazon Q Business and cross-Region inference add complexity and don't directly address the caching opportunity for repetitive inquiry patterns.
- D. Custom DynamoDB caching requires significant development effort and doesn't leverage Amazon Bedrock's native prompt caching capabilities.
Question 31
A financial services company is developing a customer service AI assistant by using Amazon Bedrock. The AI assistant must not discuss investment advice with users. The AI assistant must block harmful content, mask personally identifiable information (PII), and maintain audit trails for compliance reporting. The AI assistant must apply content filtering to both user inputs and model responses based on content sensitivity.
The company requires an Amazon Bedrock guardrail configuration that will effectively enforce policies with minimal false positives. The solution must provide multiple handling strategies for multiple types of sensitive content.
Which solution will meet these requirements?
Show answer and explanation
Correct answer: C. Configure a guardrail and set content filters to medium for harmful content. Set up denied topics for investment advice and include clear definitions and sample phrases to block. Configure sensitive information filters to mask PII in responses and to block financial information in inputs. Enable both input and output evaluations that use custom blocked messages for audits.
Option C configures a single guardrail with medium-level content filters (reducing false positives), defines denied topics for investment advice with clear samples, and critically sets sensitive information filters to mask PII in responses while blocking it in inputs, providing different handling strategies for different content types. Enabling both input and output evaluations with custom blocked messages provides audit trails for compliance. This balanced approach meets all requirements with minimal false positives.
Why the other options are wrong
- A. High severity for all categories causes excessive false positives; blocking all PII is inflexible; single guardrail lacks tiered strategies for different content types.
- B. Multiple guardrails are more complex and costly; doesn't demonstrate single guardrail can handle multiple requirements; Kendra grounding is not mentioned in requirements.
- D. Multiple guardrails with sequential chaining is operationally complex and expensive; Step Functions orchestration adds latency; this doesn't provide the unified control needed for minimal false positives.
Question 32
A hotel company wants to enhance a legacy Java-based property management system (PMS) by adding AI capabilities. The company wants to use Amazon Bedrock Knowledge Bases to provide staff with room availability information and hotel-specific details. The solution must maintain separate access controls for each hotel that the company manages. The solution must provide room availability information in near real time and must maintain consistent performance during peak usage periods.
Which solution will meet these requirements?
Show answer and explanation
Correct answer: C. Implement one Amazon Bedrock knowledge base for each hotel in a multi-account structure. Use direct data ingestion to provide real-time room availability information. Schedule regular synchronization for less critical information.
Option C implements separate Amazon Bedrock knowledge bases for each hotel in a multi-account structure, providing strong isolation and access control boundaries naturally enforced through AWS account separation. Direct data ingestion for room availability ensures near real-time updates with consistent performance during peak usage, while scheduled synchronization handles less critical information. This architecture inherently supports hotel-specific access controls and maintains performance isolation between tenants.
Why the other options are wrong
- A. A single combined knowledge base for all hotels creates data isolation and access control challenges; relying on CloudTrail filtering for access control is inadequate for true multi-tenancy.
- B. Centralizing all hotel data in a single knowledge base in a management account creates a single point of failure and makes fine-grained per-hotel access control difficult to implement reliably.
- D. A centralized agent with multiple knowledge bases still requires IAM Identity Center permission sets to enforce access control, which adds operational complexity and doesn't provide the same isolation guarantees as separate accounts.
Question 33
A company is creating a workflow to review customer-facing communications before the company sends the communications. The company uses a pre-defined message template to generate the communications and stores the communications in an Amazon S3 bucket. The workflow needs to capture a specific portion from the template and send it to an Amazon Bedrock model. The workflow must store model responses back to the original S3 bucket.
Which solution will meet these requirements?
Show answer and explanation
Correct answer: B. Create an AWS Step Functions Express workflow state machine. Use an Amazon S3 integration GetObject step to retrieve the original communications. Use an intrinsic function Pass step to parse the communications and to pass the results to an Amazon Bedrock InvokeModel step. Configure an Amazon S3 integration PutObject step to store the model responses back to the S3 bucket.
AWS Step Functions Express is purpose-built for this workflow: it retrieves the S3 object, uses intrinsic functions to parse the communication, invokes the Bedrock model, and stores results back to S3, all in a single managed workflow. This provides clear, auditable steps with native AWS service integrations and is more straightforward than Bedrock Flows (which lacks direct S3 action nodes in the parse step) or agent-based approaches that add unnecessary complexity for a defined, sequential process.
Why the other options are wrong
- A. Bedrock Flows are designed for agentic interactions rather than deterministic document workflows; the expression parsing step is less robust than Step Functions' intrinsic functions.
- C. Using an agent for a predetermined sequential workflow adds complexity; agents are designed for exploratory, multi-step decision-making rather than fixed processes.
- D. Multiple Lambda functions in a single action group is unnecessarily complex compared to Step Functions' native task integration and adds operational overhead.
Question 34
An elevator service company has developed an AI assistant application by using Amazon Bedrock. The application generates elevator maintenance recommendations to support the company's elevator technicians. The company uses Amazon Kinesis Data Streams to collect the elevator sensor data.
New regulatory rules require that a human technician must review all AI-generated recommendations. The company needs to establish human oversight workflows to review and approve AI recommendations. The company must store all human technician review decisions for audit purposes.
Which solution will meet these requirements?
Show answer and explanation
Correct answer: B. Create an AWS Step Functions workflow that has a human approval step that uses the waitForTaskToken API to pause execution. After a human technician completes a review, use an AWS Lambda function to call the SendTaskSuccess API that has the approval decision. Store all review decisions in Amazon DynamoDB.
AWS Step Functions with the waitForTaskToken API is purpose-built for human approval workflows. It pauses execution until a human completes their review, then resumes via SendTaskSuccess with the approval decision. This native workflow orchestration automatically tracks all decisions through Step Functions execution history, providing built-in audit compliance. DynamoDB stores structured approval decisions for additional auditing. Lambda functions handle the callback integration cleanly without requiring custom queue management or polling logic.
Why the other options are wrong
- A. Lambda and SQS require custom polling and queue management logic, making approval tracking less reliable than native Step Functions pause/resume semantics.
- C. AWS Glue is a data integration service, not designed for workflow orchestration or human approval tasks; it lacks appropriate pause/resume mechanisms.
- D. EventBridge with Glue jobs and ElastiCache is over-engineered for this use case and doesn't provide the structured approval workflow semantics of Step Functions.
That was 34 of 136.
The full AWS AIP-C01 pack has all 136 questions, each with the answer, the explanation and why the other options are wrong, plus a questions-only copy for timed runs. US$39, paid once, with free monthly updates and a pass-or-your-money-back guarantee.
Not ready yet? Get the first 63 free as one PDF by email. Get the Half Pack →