AWS re:Invent 2025 - Agentic AI for member-owned financials: Systems that serve (AIM337)

AWS re:Invent 2025 - Agentic AI for Member-Owned Financials: Systems that Serve

Overview

This presentation showcases a comprehensive, end-to-end solution for a member-owned financial institution's loan origination process, built using AWS Bedrock Agent Core and agentic AI principles. The solution demonstrates how member-owned financials can leverage agentic AI to streamline their loan application workflow, improve member satisfaction, and enhance the underwriting process.

Key Components of the Solution

1. AWS Bedrock Agent Core: A comprehensive agentic platform that provides the core infrastructure for deploying and managing intelligent agents, including:

   • Agent Runtime: Serverless compute for running agent applications
   • Agent Memory: Managed memory for short-term and long-term agent state
   • Agent Core Gateway: Managed MCP (Model Context Protocol) server for agent-to-agent and agent-to-tool communication
   • Agent Core Identity: Centralized authentication and authorization for agents

2. Agentic Loan Origination System:

   • The solution uses a multi-agent architecture, where specialized agents collaborate to process loan applications.
   • Key agents include:
     • Supervisor Agent: Orchestrates the overall workflow and coordinates the other agents
     • Document Agent: Validates and extracts data from loan application documents
     • Credit Risk Agent: Performs credit risk assessment using a machine learning model
     • Compliance Agent: Checks loan application for compliance with business rules

3. Agent-to-Agent (A2A) Communication:

   • The agents communicate using the A2A protocol, which enables flexible, scalable, and secure agent-to-agent interactions.
   • A2A allows the agents to discover each other, negotiate tasks, and exchange data seamlessly.

4. Agent Core Observability:

   • The solution leverages Agent Core's observability features, which provide detailed logs, metrics, and traces for the agentic system.
   • This enables end-to-end visibility and auditability of the loan origination process.

Technical Details

• The solution uses the Strands agent framework for building the agentic agents.
• Strands supports various design patterns, including A2A and workflow-based approaches.
• The agents utilize the Agent Core Gateway to communicate with external services, such as the Bedrock Data Automation service for document data extraction.
• The agents use the Code Interpreter tool to dynamically generate and execute code, allowing for flexibility in integrating with various models and services.
• Robust security measures are implemented, including OAuth authentication, authorization, and prompt injection safeguards using Strands' guardrail capabilities.

Business Impact

• The agentic loan origination system streamlines the loan application process, reducing the time and effort required to process applications.
• Improved member satisfaction through faster turnaround times and a more efficient, transparent loan application experience.
• Enhanced underwriting process by providing the underwriter with a comprehensive, auditable report on the loan application, including credit risk assessment and compliance validation.
• Increased flexibility and adaptability to changing business requirements by leveraging the prompt-based agent design.

Use Case Examples

• The solution was demonstrated with a credit union and an insurance company, showcasing its applicability to various member-owned financial institutions.
• The example highlighted the loan application process, where the agentic agents collaborated to validate documents, assess credit risk, and ensure compliance, ultimately providing a recommendation to the supervisor agent for loan approval or further review.
• The generated PDF report provided the underwriter with a detailed summary of the loan application, including all the steps taken by the agents and the final decision.

Conclusion

The AWS re:Invent 2025 presentation showcased a powerful, enterprise-grade agentic AI solution for member-owned financial institutions' loan origination processes. By leveraging AWS Bedrock Agent Core and agentic principles, the solution demonstrated how financial institutions can streamline their operations, enhance member experiences, and improve the underwriting process through intelligent automation and collaboration between specialized agents.