Providing faster answers, being available around the clock, relieving the customer center: various areas of R+V had been toying with the use of a chatbot for some time. Then, in 2018, the decision was made: The topic should be considered comprehensively and in a structured manner. "Chatbot" was launched as a trainee project.
About R+V
R+V Versicherung is one of Germany's largest insurers for private and corporate customers and belongs to the Volksbanken Raiffeisenbanken Cooperative Financial Network. R+V has been an active member of InsurLab Germany since its founding and helps shape the life of the association. Learn more
About oddball
The startup Kauz GmbH from Düsseldorf develops high-quality chatbots with natural language understanding. In addition to the NLU chatbot, the company also offers the self-developed chat software Chat & Search for internal employee service. Learn more
Every year, R+V gives graduates the opportunity to enter professional life with a trainee program. Among other things, this involves them independently tackling a cross-divisional project on a given topic. The trainees had to complete three tasks within six months: create a classification of chatbots, survey the wishes in the company and - the final challenge - create a prototype.
Step one was done quickly: write down a classification. The main requirement of the users is to get a correct answer to their question quickly. Sure, they want to make a little small talk at the beginning, but a user doesn't really want to have a conversation with a chatbot.
Step two - survey the wishes in the house: The trainees conducted dozens of interviews and first created a longlist of ideas for the use of a chatbot. These use cases ranged from customer support to marketing scenarios to internal services for employees. The longlist became a shortlist, and in the end, R+V's internal technical support - the Central User Service - emerged as the winner and was selected as the use case for the prototype. An important reason for this choice was the existing self-help content, which was not well received.
Last task: quickly realize a working chatbot, because the time of the project was approaching the end.
And that's where "Kauz" came into play. A "no-code solution" that could be implemented directly by the interdisciplinary trainee team quickly and without major IT effort. Based on the realization that users want a correct answer to their question immediately, the team focused on the "Chat&Search" component.
Just six days later, the text-based chatbot was flickering and was christened Lara. The first test run produced sensational results without adjustment, with a hit rate of 77 percent correct answers compared to the intranet search, which achieved a hit rate of 8 percent.
Mission completed - and now?
The Central User Service took over the prototype as a "digital apprentice", who was then also given an "instructor". Successively, Lara was further trained with the information already available. Lara was able to respond to technical questions by either helping people to help themselves or by picking up other topics and passing them on.
The trainees already had it in mind that Lara's skills would be expanded. Later, she was to provide general support to R+V employees. That's why they had already fed Lara other facts and figures from the company in addition to the technical content. She also knows abbreviations.
In 2020, the content originally entered for technical support was already less than half. Lara has made good progress on the way to becoming the digital R+V assistant for all employees.
And she was wrong with her answer in only 8.4 percent of the questions. In 2021, the expansion continues. Some new topics were already added at the beginning of this year and more are in the queue.
Conclusion: "Kauz" is a "no-code" solution for quick and easy use by departments. It is easy to maintain and further develop in terms of content. Because it is a curated chatbot, the development team does not have to worry about language understanding and small talk content.
