Learn how EXL helped insurers reduce the cost of underwriting by 80 percent with a generative AI–based virtual assistant built on Amazon Bedrock.

Benefits/Key Metrics

• 80% – reduction in insurance underwriting costs
• 60 days – to build generative AI application instead of many months
• From Days to hours – to complete underwriting processes

Opportunity | Reducing Manual Effort in Underwriting Processes

Data analytics and digital solutions company EXL offers insurers a business and underwriting digital platform called the Life Digital Suite (LDS™) that streamlines and automates new business to policy-issue processes. With its flexible app and workflow capabilities, LDS™ simplifies case management, provides valuable data insights, and makes product launches easier for insurers and their customers; companies have used LDS™ to support the launch of over 100 new products across eight countries.

At the same time, much of the underwriting process remained manual. Kamal Deep Singh, senior assistant vice president of EXL, explains, “Underwriters were still spending days reviewing hundreds of pages of documents as part of the initial assessment and review stages.” EXL sought to reduce the workload on underwriters by using generative AI to process documents and deliver faster, more accurate evaluations. However, it needed to address several technical challenges. Singh explains, “We knew that generative AI hallucinations could affect the reliability of the results, and we had to maintain the confidentiality and integrity of all data, including healthcare, in compliance with regulations like PII [Protected Identifiable Information] in India.”

Solution | Developing Generative AI Virtual Assistant for Underwriting

EXL chose to work with AWS to build a generative AI–based solution that could overcome these technical challenges. Shortly after beginning its collaboration with AWS in 2022, EXL developed EXL Customer 360 Insights, a solution that consolidates customer journeys across multiple data sources. The solution makes use of Amazon Bedrock, a fully managed generative AI service. Says Singh, “Amazon Bedrock offers access to multiple LLMs [large language models] to do extensive testing to reduce the risk of hallucinations for our specific use case.”

Moreover, AWS provides a centralized platform for EXL’s AI needs, streamlining development and enhancing the capabilities of its generative AI offering. “The strong security focus at AWS also makes it easier for us to comply with regulations such as PII here in India,” says Singh. The EXL digital team experimented with various LLMs and foundational models, including Mistral AI, Claude 2, Sonnet 3, and Amazon Titan. Working closely with Amazon Solutions Architects and AWS teams further expedited development. “AWS provided clear guidance and support,” Singh recalls, “which helped us move efficiently through the development process.”

Next, EXL developed LDS™ Underwriting Assist, a retrieval-augmented generation chatbot that helps insurers streamline assessment and review stages. Integrated into the LDS™ platform suite, the service uses Sonnet 3 through Amazon Bedrock and employs Amazon Kendra as the chatbot interface. To simplify deployments for customers, EXL built a custom API layer with FastAPI and containerization. The solution also integrates Amazon Textract to extract data from scanned documents and Amazon Comprehend to analyze text, both of which offer features to redact PII and protected health information to ensure data privacy and compliance.

EXL launched the LDS™ Underwriting Assist service in August 2024, just 60 days after starting development. “With the right tools and expert support from AWS, we significantly reduced development time,” says Singh. “This also gave us the competitive advantage of introducing such a powerful tool to the market ahead of other companies.” Not only does EXL benefit from its new platform’s reliability and scalability, but it has also optimized costs by paying for Amazon Bedrock and other AWS services only when customers use the chatbot solution, plus avoids the expense of managing the underlying infrastructure.

Outcome | Reducing Underwriting Time and Costs with Amazon Bedrock

Using the AWS-based generative AI solution, insurers can reduce the underwriting process from several days to a few hours. Based on EXL analysis, the solution also has the potential to save insurers up to 80 percent of underwriting costs. “We’ve created a generative AI solution with AWS that significantly increases the efficiency of insurance underwriting while dramatically lowering the associated costs,” says Singh. Several existing LDS™ customers onboarded the assistant service in the first couple of months after it went live. Ajmal Malik, product manager of LDS™ at EXL, adds, “We are excited to work with AWS to deliver this innovative solution to the market. This approach improves underwriting time, enhances accuracy, and reduces costs in the process.” The growing adoption has also driven an increase in the number of deals signed with customers.

EXL plans to explore more generative AI use cases, building on its expertise with Amazon Bedrock and other AWS AI services. “The opportunities for generative AI are growing rapidly worldwide,” Singh concludes. “While many companies are still navigating the technology to bring their ideas to life, we’ve moved beyond that stage with AWS. We can now confidently use this technology to deliver better customer service.”