Amazon Bedrock AgentCore automates first notice of loss for insurance claims

Insurance companies spend huge amounts of time on routine data entry when they first register a claim. AWS has published the architecture of a system that removes this work entirely: an AI agent interacts with insurers' web portals on its own, with no operator involved.

What FNOL is and what the problem is

FNOL (First Notice of Loss) is the mandatory first step in any insurance case: a car accident, a fire, a medical incident or the theft of property. An operator takes the customer's call, collects the details of the incident and enters them into the insurer's systems.

Most of this work is strictly mechanical. The specialist copies data from one window to another, fills in dozens of standard fields, checks the policy number and switches between several systems. Errors in manual entry cause processing delays and repeat calls to customers. At large insurers, FNOL consumes tens of thousands of person-hours every year.

How the system works

AWS proposes a solution built on two key components:

• Strands Agents SDK handles logic and reasoning. The agent receives information about the loss, analyzes the context, classifies the type of case and determines the sequence of actions according to the specific insurer's rules.
• Amazon Bedrock AgentCore Browser Tool does the work in the browser: it opens insurance portals, fills in forms, clicks buttons and reads the results, just as a human operator would.
• The system works with real web interfaces without additional API integrations.
• A new portal can be connected without development work; the agent figures out the interface on its own.
• A human steps in only for complex or non-standard cases.

The fundamental difference from the traditional approach is the abandonment of custom integrations. Automating insurance processes usually requires building a separate connector for each system. Browser Tool perceives the screen as a human does and controls the browser directly, which removes that dependency.

The browser-agent approach is not limited to insurance: banking, government services and medical records face the same problem of manual entry into web portals.

Strands SDK and the role of the specialist

Strands Agents SDK is AWS's open-source framework for building AI agents. In insurance automation, it is responsible for the "thinking" part of the job: understanding what happened, which data is needed, whether the details contradict each other and whether the case meets the policy terms. The agent works iteratively: it asks clarifying questions, verifies data and requests additional information when necessary.

Strands SDK is distributed as open source, so companies can add their own processing rules or integrate it with existing decision-making systems.

The system does not replace insurance experts; it takes over the strictly mechanical work. Complex situations that require professional judgment, such as potential fraud, non-standard losses and disputed circumstances, remain the responsibility of humans.

"The approach preserves human expertise while eliminating repetitive

screen work," is how AWS describes the principle behind the system.

Freed from routine, specialists can focus on cases where experience is genuinely needed: settling complex claims, dealing with customers in stressful situations and spotting suspicious patterns in the stream of claims.

What this means

Browser-based AI agents are becoming a working tool for back-office automation at insurance and financial companies. The AWS architecture built on Strands SDK and Bedrock AgentCore Browser Tool sets a template that adapts to any portal without rewriting integrations.

For the insurance industry, this means a lower barrier to adoption: instead of multi-month integration projects, it is enough to configure the agent for a specific web interface.