Transforming Insurance with AI
This article was co-authored by Andrea Anderson (Head of User & Information Experience) and Nada El Maliki (Instructional Designer)
At Guidewire, we always seek new ways to innovate and stay ahead of the technology curve. Our recent AI Science Fair showed our dedication to exploring the potential of AI technologies and how they might be used to revolutionize our industry and best serve our customers and end-users.
This virtual event brought together over 300 employees from a variety of business backgrounds, including QA testers, Engineers, Consultants, Writers, User Experience Designers, Business Analysts, Marketing, and Finance people, all with the goal of exploring the potential of generative AI technologies and large language models and how they might add value to Guidewire’s products and operations.
We gave teams a simple mission: Get hands-on experience with large language models and imagine how these models could benefit Guidewire as a company, and help us create valuable solutions for our customers. We didn’t ask for code or models. Instead, we encouraged finding and making solutions utilizing natural language processing concepts and AI tools, in areas such as knowledge sharing and guidance, code generation and review, and interactive experiences.
The event was a resounding success, with participants showcasing innovative ideas and how AI can transform the insurance industry, as well as help us run better as a company.
Over 100 ideas were generated, with 57 projects submitted across various topics ranging from QA & Testing, insurance business configuration, end user assistance, knowledge and education, and code making.
Here are some highlights:
Concept: WiseMonkey
“WiseMonkey” generates a “strawman” insurance product model by harvesting data from policy documents such as a product fact sheet, a policy contract, policy wording, and the like, and creating the product mindmap. The team envisioned a solution that could be used by insurance agents to quickly analyze and understand complex policy documents, allowing them to better serve their customers. The application uses natural language processing (NLP) to extract information, resulting in an Advanced Product Designer (APD) Mind Map.
Concept: Guidewire claims assistant
The Guidewire claims assistant makes filing claims conversational and flexible, allowing claimants to provide claim details in any order. It features a flexible conversational user interface that guides the claimant through the process with the help of smart AI (ChatGPT). The claims assistant interprets the claimant’s responses and prompts them for missing information. The claimant can review and edit responses, and receive real-time feedback on claims processing.
Claimants get personalized, empathetic guidance and real-time feedback, making the process of filing their claim less stressful and more efficient. It also frees up insurer CSRs to focus on helping claimants further in the claim lifecycle, and allows claims teams to spend less time on gathering the right data up-front for quicker claims resolution.
Concept: Guidewire AI-powered training platform
Creating software training is hard and time consuming for content creators because they are not the product experts. We used openAI to create new learning content based on product documentation and unstructured knowledge sources. We then took this content and used it as the basis for synthetic AI video that mimics a human speaker. These AI videos can be produced in high volume and in a way that is easily translated into multiple languages at a low cost. With the availability of APIs for artificial intelligence platforms, software tooling can be coded to automate production of media at scale. We used Synthesia to generate the learning modules.
Concept: Use NLP to generate tests for globalization testing
Developing test cases can be challenging for developers who may not be familiar with region specific formats (for example, date/time, currency, address, telephone, calendar, sorting, etc).
Developers need to manually identify complicated region-specific formats for many product lines and come up with release-specific test cases, which can be very time consuming and costly.
The team who worked on this project generated tests for globalization testing by using NLP models. This solution would assist development teams in generating specific formats for a given region at runtime and executing globalization testing in a more streamlined way. Better tests increase coverage for globalization testing, which can be done for multiple product lines and in different regions.
Concept: Jutro Smart CodeMod
Design systems and UI frameworks constantly evolve, and sometimes breaking code changes cannot be avoided. So how do you keep your digital application on the latest codeline with minimum effort?
Jutro (Guidewire’s design system & UI framework) Smart Codemod is trained with Jutro’s react-based code, and understands the context of the target application to be upgraded, allowing for sophisticated and complex code replacements in Jutro Apps when breaking changes are introduced. Our hope is to make upgrading to the latest and greatest UI library a snap.
What’s next?
The AI Science Fair has been a significant milestone for Guidewire. It has allowed us to push the boundaries of what is possible in the P&C insurance industry. It was also a powerful reminder of the importance of fostering a culture of innovation within large organizations where employees are encouraged to think outside the box and take calculated risks to create something truly groundbreaking.
As we think about the future of AI at Guidewire, there are many questions that we will need to answer. For example:
- What boundaries should we be aware of as we explore potential applications of AI to the P&C industry?
- What do we need to understand in order for AI to be used in an ethical and fair way?
- How do we continuously learn and iterate on this topic?
- How do we identify use cases and personas to leverage AI for?
- How do we determine our preferred tools and reference architecture?
The AI Science Fair provided a great starting point for these types of conversations, and we look forward to continued dialogue and collaboration on the topic of AI.
Note: This article was authored and written by humans.
