A national insurance coordination organization deployed an AI agent using Amazon Bedrock and Amazon Lex to modernize customer service operations. The solution provides 24/7 customer support while reducing routine inquiry handling costs by 40%.
Customer Background
This organization coordinates insurance companies offering motor third-party liability insurance in a European market. As the national bureau representing their country in the international insurance system, they serve as the critical link between insurance companies, regulatory bodies, and the broader automotive insurance ecosystem.
Challenge
The organization faced mounting pressure from high volumes of repetitive customer inquiries that were overwhelming support staff and limited service availability to business hours. Customers struggled to find relevant policy information and navigate complex insurance procedures. Without modernization, the organization risked losing customers to competitors with superior digital experiences while facing potential compliance issues in the highly regulated insurance industry.
Solution
Trek2Summit developed a comprehensive customer service AI agent using:
- Amazon Bedrock - Using Claude 3.5 Sonnet for superior reasoning capabilities in handling complex insurance regulations
- Amazon Bedrock Knowledge Bases - Integrated with insurance policy documents, regulatory guidelines, and procedural manuals
- Amazon RDS Aurora with pgvector - Vector database for RAG implementation
- Amazon Lex - Conversational AI interface
- Amazon Lex Web UI - Web application frontend
- Amazon S3 - Knowledge base document storage
The solution was deployed across multiple availability zones in the Frankfurt region with RDS read replicas for database scaling. Extensive prompt engineering ensured compliance with insurance regulations and accurate policy interpretations while maintaining professional industry terminology.
Results
- 70% of customer inquiries resolved without human intervention
- 40% reduction in routine inquiry handling costs
- 24/7 customer support availability achieved