AWS and Strands automate claims intake with hands-free AI agent system
AI agents handle portal navigation and evidence interpretation, freeing adjusters for higher-value decisions.
Manual first notice of loss (FNOL) processing consumes significant adjuster time on repetitive tasks: navigating portals, verifying evidence completeness, and interpreting unstructured artifacts like photos, videos, and scanned documents. During volume spikes from catastrophes, these delays compound, creating backlogs. The complexity lies in correlating multimodal data before meaningful assessment can begin—work that currently demands expert judgment applied through slow screen-by-screen clicks.
The solution combines Strands Agents (open-source SDK for insurance-specific business rules) with Amazon Bedrock AgentCore Browser Tool and Nova Act for browser reasoning. Nova Act interprets natural-language instructions to perform portal actions (e.g., 'open next claim'), while Strands agents apply domain logic for evidence interpretation and claim complexity assessment. The result: claims professionals receive pre-tagged, decision-ready submissions, eliminating manual validation while maintaining auditability and human oversight.
- Strands Agents SDK applies domain-specific business rules (evidence correlation, coverage assessment) using foundation models on Amazon Bedrock.
- Nova Act drives portal interactions via natural-language instructions, reducing adjuster screen time by automating repetitive navigation and validation.
- The workflow yields tagged, structured evidence as a reusable asset for routing, pattern analysis, and continuous claims workflow optimization.
Why It Matters
Reduces claims processing bottlenecks by automating repetitive intake tasks, enabling faster resolution during catastrophe surges without sacrificing expertise.