AWS re:Invent 2025 - Agentic AI Solutions for Financial Services & Insurance with AWS (PEX301)

Agentic AI Solutions for Financial Services & Insurance with AWS

Challenges in Insurance Underwriting

• Insurance underwriters currently rely on a largely manual process to assess risk and determine premiums
• They must gather and synthesize data from 15-20 different systems to evaluate factors like property details, historical claims, weather patterns, etc.
• This time-consuming data gathering and analysis leaves underwriters with little time to apply their expertise and make informed decisions
• As a result, insurers are only able to process a fraction of submissions, leading to lost customers

Key Strategies for Insurers

1. Granular, location-based pricing instead of broad zip code-based pricing
2. Coverage modifications like excluding certain risks (e.g. wildfire, flood)
3. Exiting high-risk markets where they cannot provide adequate protection
4. Providing remediation recommendations to customers with premium discounts

Agentic AI Solution Overview

• Utilizes a "swarm" design pattern with multiple autonomous AI agents collaborating to gather and analyze data
• Agents handle tasks like:
  • Geocoding addresses
  • Assessing weather/climate risks
  • Analyzing historical claims data
  • Synthesizing risk factors
  • Providing underwriting recommendations
• Agents leverage tools and data sources like AWS Location Service, NOAA weather data, and insurance claims databases
• The underwriter reviews the comprehensive risk assessment provided by the agent swarm and makes the final underwriting decision

Technical Implementation

Strands Agents

• Open-source Python SDK for building AI agents with just a few lines of code
• Leverages state-of-the-art language models (including Amazon Bedrock models) for reasoning, planning, and task execution
• Provides pre-built tools for common agent use cases like knowledge management, web search, and API calls
• Supports custom tool development and integration with the Model Context Protocol (MCP)
• Includes built-in observability and traceability features

Amazon Bedrock Agent Core

• Suite of managed services to simplify the deployment and operation of production-ready AI agents
• Agent Core Runtime: Secure, serverless runtime with true session isolation and support for long-running workloads
• Agent Core Identity: Manages identity and access for agents to access AWS and 3rd party resources
• Agent Core Gateway: Enables easy integration of existing APIs and services as agent tools
• Agent Core Memory: Fully managed service to maintain short and long-term conversational context
• Agent Core Observability: Provides visibility into the end-to-end agent execution trace

Business Impact

• Transforms the insurance underwriting process from manual and time-consuming to automated and efficient
• Enables underwriters to focus on applying their expertise rather than data gathering and synthesis
• Helps insurers expand their risk coverage, provide more personalized pricing, and better serve customers
• Allows for faster, more informed underwriting decisions leading to reduced losses and improved profitability

Example Use Case

• Underwriter receives a submission for a commercial property in Houston, TX
• AI agent swarm automatically gathers and analyzes data on the property, location, weather/climate risks, and historical claims
• Agents provide a comprehensive risk assessment, including a recommended premium structure
• Underwriter reviews the agent-generated report and makes a conditional approval decision, requiring the customer to implement certain risk mitigation measures

Key Takeaways

• Agentic AI solutions can transform complex, data-intensive business processes like insurance underwriting
• Strands Agents and Amazon Bedrock Agent Core provide a powerful platform to build and deploy production-ready AI agents
• Collaborative "swarm" of agents can gather, analyze, and synthesize data to support human decision-makers
• Managed services in Agent Core eliminate the operational complexity of running AI agents at scale
• Adoption of these technologies can help insurers improve risk assessment, expand coverage, and better serve customers