Engineering and user experience of chatbots in the context of damage recording for insurance companies

Abstract

With ever increasing amounts of data and changing consumer expectations, the German insurance sector is undergoing immense transformation. While insurance has traditionally been an industry with very low customer engagement, insurers now face a young generation of consumers who expect quick and on-demand services at a time suitable for them. For this reason, specialized digital assistants (chatbots) could become a first line of support, driven by rapid advancements in artificial intelligence as well as vigorous growth in the adoption of messaging services. They promise to be scalable, accessible around the clock, and to improve customer engagement by orders of magnitude as opposed to traditional channels such as email or telephone.

Another key issue is that insurance claims are currently touched by multiple employees in a process referred to as the traditional workflow. In order for insurance companies to remain competitive and become truly forward-leaning carriers, they need to reduce their loss adjustment expenses (LAE) while delivering customer-pleasing solutions in a digital age. Given the prospect of modern data analysis capabilities and conversational user interfaces (CUIs), claims processing in the future may be performed without any human intervention at all, which is known as the virtual or, most desirably, touchless handling.

First explorations have shown that chatbots are indeed a feasible solution to the aforementioned problems, however they indicated several issues of frustration in terms of user experience, resulting in a direct impact on user satisfaction and task success. Therefore, this thesis aims to explore the constituents of an engaging and satisfying conversational interface and ways to achieve a better user experience.

In an ongoing project on innovation management in collaboration with insurances, the industry partner Fraunhofer IAO continuously explores ways to advance the business processes of their customers. For this reason, they commissioned a prototypical implementation of a chatbot for smartphone incident recording as part of this examination. This high-fidelity prototype should be able to record cases of damage in a convenient manner, and ultimately serve as a showpiece to inspire confidence in German insurers that employing chatbots in their workflows would be a viable asset to target increasing consumer expectations, improve engagement, and cost-effectively scale their customer support.

As the details of the eventual solution were not prescribed in any way, it was possible to be more experimental in terms of utilized technologies and a methodical approach to requirement analysis. Hence, the technical and human-centric requirements, which deemed useful to realize in order to achieve the goal of a satisfying user experience, were drawn from systematic review and analysis of existing academic literature and best practices from the industry. The implemented requirements are thus presented in their appropriate chapters, followed by challenges and design decisions, as well as steps undertaken to assure software quality in past and future iterations of engineering on the prototype. Thereupon, in addition to an assessment of the functional requirements, a survey on the basis of quality attributes was conducted to evaluate the fulfillment of soft requirements.