Building a Conversational Data Collection Agent on Amazon Quick

This article demonstrates how to build a conversational data collection agent using Amazon Quick and Bedrock AgentCore Gateway to replace manual spreadsheet-based questionnaires for structured data collection from distributed stakeholders.

• Stakeholders complete structured questionnaires through natural conversation with multimodal input support (text, diagrams, wiki pages)
• User-authenticated MCP tools enforce per-user data isolation and row-level access control via Amazon Cognito JWT tokens
• Responses are persisted in Amazon DynamoDB with real-time cross-answer validation to detect contradictions
• Automated summarization pipeline transforms free-text responses into structured reports using Amazon Bedrock with domain rules stored as S3 Markdown files
• LLM-as-a-Judge evaluation layer scores each report across extraction accuracy, rule compliance, and hallucination detection
• Architecture is domain-agnostic; cloud migration discovery is the example use case but applies to compliance assessments, vendor onboarding, and security reviews

The solution combines Amazon Quick's conversational interface with Bedrock AgentCore Gateway's MCP tool orchestration to automate discovery, summarization, and evaluation phases of structured data collection at enterprise scale.