A Leading Healthcare BPO Transforms Insurance Policy Document Creation with Rapyder’s Cutting-Edge GenAI Solution

Introduction:

Customer, is a global leader in business process management (BPM), optimizing the member/patient experience within the healthcare industry. Cutomer is a tech-enabled BPM services provider and a trusted thought partner offering a wide range of transformational services to enhance efficiency and quality of care across the healthcare system.

Client:

A Healthcare Customer

Industry:

Healthcare

Offering:

Gen AI

AWS Services:

1. Amazon EC2
2. Amazon Bedrock (Claude Anthropic Models)
3. Amazon S3, CloudWatch
4. AWS Lambda
5. Amazon API Gateway

Business Need:

The customer requires a streamlined, efficient solution to accelerate insurance claim processing. They are seeking an automated system capable of generating multiple-choice questions and flowcharts from insurance policy documents. This solution aims to significantly reduce processing time while improving the accuracy and consistency of claim assessments.   

To address this issue, they sought an automated solution capable of generating these Q&A documents by summarizing the content of the policies based on predefined workflows.

Implementation:

To meet the customer’s requirements for reducing processing time and improving accuracy in creating the Q&A documents, Rapyder Cloud Solutions has developed a comprehensive and innovative solution leveraging generative AI and AWS services. Our solution not only automates the entire insurance claim processing process but also incorporates several key features to enhance efficiency and accuracy. 

1. A robust data pipeline is built using Amazon API Gateway, Amazon S3, AWS Lambda, and Amazon Bedrock to facilitate seamless communication and data flow between various components of the system.
2. Utilization of Bedrock’s Anthropic Model, specifically Claude 2.1, for processing policy documents. The choice of Claude 2.1 underscores the emphasis on accuracy, which is crucial for generating precise Q&A documents.
3. Integrated Prompt engineering and a rule-based approach to create the Q&A document. This allows for flexibility in extracting information from policy documents, either through static rules or custom queries.
4. Incorporation of a human feedback approach and dynamic rules as part of the pipeline to enhance and improve the quality of the Q&A documents. This ensures that domain-specific expertise is integrated into the process, resulting in more accurate and relevant outputs.
5. The Streamlit web app features an intuitive interface for easy navigation and enhanced productivity.
6. Outputs in Excel and JSON formats are enabled for easy integration with existing systems, ensuring accessibility and usability.

LLM-based Key Work and Improvements:

• Model evaluation and change are facilitated using Amazon Bedrock Claude 2.1 for summarizing policy documents and preparing QnA.
• Implemented an In-Context learning approach to gather feedback and input from Healthcare TPA and specialists, enhancing the quality of the document.

Reaping Rewards:

• The Q&A and Policy assessment document creation that traditionally took 8-12 weeks, with the solution, the turnaround time is reduced to 1 week, impacting both timeline and operational cost.
• The solution achieves a 70% accuracy in processing documents and generating the final Q&A document. 
• The automated Q&A document creation solution has been successfully deployed across the customer’s operations, resulting in improved resource utilization and an overall 70% reduction in manual intervention efforts.
• As the resources previously involved in this activity have been reallocated to other tasks, customer was able to save costs upto $1million. The solution is designed to benefit more than 150 business accounts across multiple continents.
• The option of human in loop for collecting feedback from the specialist has helped improve the accuracy of the solution.