Extending public sector intelligence with Agentforce and AWS

This article demonstrates how to combine Amazon Bedrock Data Automation with the Model Context Protocol to process unstructured evidence in public sector agencies and surface insights through Salesforce Agentforce.

• Event-driven architecture processes multimodal data (documents, images, video, audio) using Amazon Bedrock Data Automation
• Amazon Bedrock AgentCore Gateway exposes processed data as MCP tools for Salesforce Agentforce to discover and invoke
• Unstructured files stored in Amazon S3 trigger Lambda functions that extract structured insights and store results in DynamoDB
• Salesforce users query processed evidence through natural language in Agentforce without switching systems
• Modular pattern extends beyond evidence management to permits, benefits claims, and compliance reviews
• Security considerations include input validation, MCP allowlist scoping, and Amazon Bedrock Guardrails for content filtering

The architecture enables public sector agencies to automatically extract insights from unstructured evidence and make them accessible through natural language queries in Agentforce, with a reusable pattern applicable to other document-heavy workflows.