Interview with Isabella Martorell Nassl, Versicherungskammer

MARTORELL NASSL / As with all companies, it is of the utmost importance for us to know our data inside out. The better we can surmise the needs of our customers based on the data, the more focused we can be in responding to those customers. Furthermore, the Versicherungskammer Group is a growing company with an ever-increasing premium volume. We are therefore constantly seeking to optimize processes. Managing the additional volume of work efficiently means we need to automate processes. That is how we came upon the Watson cognitive system. As we focused more closely on technology, we thought: “We’d love to try that out in our company!”

MARTORELL NASSL / We use cognitive systems to detect discontentment and a desire for certain offerings among our customers. We receive approximately 20,000 letters each day, all of which are scanned, digitized, and registered. Finally, we use cognitive systems to check whether these documents contain expressions of discontentment or for certain specific requests. If they do, we mark them explicitly and they are forwarded automatically to the relevant employee. This enables us to act in a more targeted manner, optimize our service and thereby increase customer satisfaction. Furthermore, cognitive systems free up our administrative staff, who then have more time for individual topics – in keeping with the motto “Individualization in the age of automation”.

MARTORELL NASSL / I’ll need to give you a bit of background to answer that question. In order to find out whether Watson even worked the way we imagined, we launched a pilot project in 2015 together with IBM and Munich University of Applied Sciences. It was so successful that we actually won an award for our pioneering work in the insurance industry. So we began implementing the system in March 2016 and went live after nine months of preparation. When configuring the system, we focused particularly on ensuring that it would be able to communicate with our existing systems. We also took a close look at the follow-on processes, adapting and optimizing these fairly quickly with the help of our own employees.

MARTORELL NASSL / Do you know how a cognitive system learns? We always compare it with a three-year-old child. It can do the basics, but we need to send it to school – and ideally also to university at some point. We began the learning process with a number of letters and started Watson on a cycle of reading and understanding them. To get to that point, administrative workers first classified around 1,000 documents and transferred them to the system. Afterwards, we checked what Watson could detect, and we then made manual adjustments. By the end of the project, we had fed around 40,000 documents into the system – and Watson continues to learn from them. At the moment, Watson is completing the equivalent of an apprenticeship, but we aren’t far off adding a Bachelor’s degree.

MARTORELL NASSL / As always with new things, you first need to persuade the people involved. To this end, it was essential to get both employees and management involved from the outset, and allay any fears they may have. That was the greatest challenge, and we overcame it through open communication and a very transparent process.

MARTORELL NASSL / That is a difficult question. We have had such positive experiences to date that we definitely want to continue working in this area. At the moment, we are considering a great many areas of application. For example, it is easy to imagine strengthening our knowledge management using cognitive systems. However, that is just an initial idea. We are currently in a start-up phase, during which we will first collect, assess and then decide.

MARTORELL NASSL / I personally think that cognitive systems and artificial intelligence can be used anywhere there are large quantities of unstructured data, the reading and evaluation of which involves a substantial effort. This can happen, for example, as part of recommended actions for experts. If supported by artificial intelligence, professionals would be faster and, thanks to shorter training periods, would also be in a position to make recommendations. I can easily believe that the topic of speech will keep coming up, well into the future. We at the Versicherungskammer Group are already examining approaches to determine how cognitive skills can be integrated into speech.

For more information on the Versicherungskammer group, please see www.vkb.de.