The client is a leading independent provider of claims management solutions.
The Challenge
Ensuring compliance and consistency across large numbers of claims is a significant challenge for insurance adjusters, especially when dealing with complex cases. Adjusters must sift through numerous documents and verify multiple details, and access and analyze various policy documents to make informed decisions during the claims resolution process. Manually locating and interpreting relevant information can be time-consuming and stressful, fraught with the potential for errors. The client – a global provider of claims management solutions in 70 countries – needed a solution that would reduce the time and effort required for claims processing, improve efficiency and customer satisfaction, and cut costs.
The Solution
Persistent developed a cloud-based LLM reusable framework that chained together different components to create “talk-to-policy” documents, generative Q&A, and summarizations. A primary goal of the Generative AI solution was contextual understanding, i.e., gleaning context from whatever information was being considered, thereby ensuring accurate interpretation of policy terms and conditions.
The scalable, serverless LLM integration was built on AWS Bedrock, a fully managed service that offers a choice of high-performing foundational models from leading AI companies. Qdrant Vector Database was employed on the back end for advanced vector similarity search technology. The solution also leveraged LangChain’s use cases for document analysis and summarization, plus HF embeddings, which are numerical representations of pieces of information, e.g., texts, documents, and images.
Finally, as a Premier Tier Services Partner in the AWS Partner Network, Persistent contributed its extensive AWS expertise to this use case, based on our practice knowledge from working with clients to design, architect, migrate, modernize, and manage their AWS cloud workloads.
The Outcome
The AWS-based Persistent solution has delivered enhanced insights from the massive flow of documents that adjusters employ in their work and enables adjusters to interact with policy schedules and documents using natural language. Overall, “process automation” has taken the place of manual effort, as GenAI automates claims assessment tasks by swiftly reducing time-consuming and mundane checks and errors. The bottom line is greater efficiency and significant cost savings for the client.
Technology used:
- AWS Bedrock
- LangChain
- Qdrant Vector DB
- HF embeddings
- Python
- React.js
