Insurance claims following major disaster events are often delayed due to documentation gaps, slow damage assessments, and policyholder confusion. This presentation introduces a technical solution that leverages open-source satellite imagery and Large Language Models (LLMs) to enhance the customer experience and expedite the claims process. Using Synthetic Aperture Radar (SAR) and multispectral imagery, this tool performs automated change detection to preliminarily identify disaster-related damage ahead of field inspections by responders and insurance adjusters. These results are integrated into a conversational LLM assistant trained on relevant FEMA documentation to provide incident- and property-specific courses of actions to survivors beginning to navigate the claims process. A dashboard interface enables survivors to visualize and interact with preliminary damage results, prompt the LLM assistant for claims information, and export relevant information to bring to their community and insurance company. We will discuss the open-source data, geospatial analysis, and LLMs used to develop the claims assistant tool. We will then present use cases with historical incident data to demonstrate how the outputs of our tool can inform policyholders' claim experience. Through this presentation, we will show how policyholders claim experience can be improved using open-source data, imagery change detection, and LLM assistants.
