Insurance AI Agent

Insurance AI Agent

Cornell University - Ithaca, NY

Overview

Our overall goal for this project is to create a data-driven, proof-of-concept prototype of the claim filing experience using a native mobile application integrated with Generative AI and virtual agent technologies. More specifically, our design objectives were to optimize the claim filing user flows, elevate the success rate of filing claims, and improve the overall user experience. Complementing these user-centric goals, our business objectives aimed to boost the percentage of users filing claims online while reducing the costs of phone and live chat customer service.

Tools

  • Figma

  • Miro

TEAM

  • Doreen Luo - Project Manager

  • River Yu - Client Coordinator

  • Charon Zhou - UI Designer

  • Yuning Zhang - Researcher

  • Yining Wang - UI Designer

Timeline

Start: Aug 2023

End: Dec 2023

Research

Our project's research phase was pivotal, involving thematic analysis of over 1,500 survey entries to understand user pain points and expectations. Top issues included a lack of clear communication on claim denials, unreliable on-site repair services, and problematic customer service interactions. Users also expressed frustration with the complex claims process, redundant data entry, and lack of clear guidance and feedback. This in-depth analysis informed our user-centric design, ensuring our solution directly addressed these core concerns to enhance the overall user experience.​​

Low Fidelity Prototype

The low-fidelity prototype phase of our project was integral in shaping the direction of our final product. It served as a tangible representation of our solution, crafted to address the various user pain points unearthed during our research. This prototype phase focused on refining the claims process to make it more user-friendly, with an emphasis on developing a conversational AI that could navigate users with clear and intuitive guidance. It laid the groundwork for the high-fidelity prototype, providing a visual and interactive means to test and iterate on our design concepts effectively.​

Usability Testing

Our first usability test involved 13 participants, each with experience in at least one type of insurance. The test employed a think-aloud protocol, where participants vocalized their thoughts while interacting with the prototype on mobile phones. This approach provided valuable insights into the user experience and highlighted areas for improvement. We discovered that the prototype's heavy reliance on user typing, instead of selection, increased user effort and detracted from the intuitive experience we aimed to offer. Additionally, the output style didn't closely resemble that of a generative AI, leading to information overload. Some components inherited from existing desktop prototypes also caused user confusion due to their interaction designs. These findings were crucial in guiding the subsequent iterations of our prototype, ensuring a more streamlined and user-friendly interface​

High-Fidelity Prototype & Design Systems

In our high-fidelity prototype, we concentrated on refining the user experience and interface design. Key improvements included resolving visual issues, adding an edit button under users' messages, and introducing a menu bar for better navigation. We emphasized giving users choices through AI prompts, reducing the need for manual typing, and providing specific details like policy rates to enhance clarity. The design system focused on visual design elements that reinforced branding consistency, professionalism, trustworthiness, and a user-friendly and accessible interface, aligning with our goal of creating a seamless and intuitive user experience.

Solution Showcase:

Our AI chat interface optimizes the service experience, balancing client needs with user convenience. The solution features a chatbot that streamlines the claim process by guiding users through an intuitive self-diagnosis and offering personalized service recommendations. Improvements based on user feedback include AI-assisted selection prompts and a streamlined design for easier navigation. The interface boasts an 'edit' function for correcting inputs, a toggle for light and dark mode, and multilingual support. User testing confirmed the interface's improved intuitiveness and efficiency, validating the design's effectiveness in enhancing user engagement and satisfaction.

Due to the project client’s requirement, the Logo and Firm Name are replaced with placeholders in the prototype.