“Knots” must be untangled

InnoNetz — Artificial intelligence for process automation in the “Innovationsnetzwerk Digitalisierung für Versicherungen” (Digitalization for Insurance Companies Innovation Network)

© Fraunhofer IAO, Illustration: Thomas Kuhlenbeck

Process optimization helps insurers become more cost-efficient and customer-friendly. Insurance companies are researching the role of artificial intelligence in this optimization with Fraunhofer IAO in this joint innovation network.

The challenge

When Verena Pohl and Dr. Maximilien Kintz are struggling with tangled balls of “spaghetti,” they are not necessarily at the dinner table. “What we call spaghetti diagrams,” says Pohl, “are process diagrams where you have to use intertwining lines to understand how the processes in an insurance company work – starting from receiving a customer email, for example.” Like her colleague Dr. Kintz, she has long been active on the Fraunhofer IAO side of the “Digitalization for Insurance Companies” innovation network, in which a number of well-known insurance companies exchange ideas on future-oriented topics.

But it’s the many knots in the “spaghetti” that are the problem. Processes developed over long periods of time in insurance companies are often laborious, redundant or, in the worst cases, simply chaotic. As a result, customers have to wait too long for communications, inquiries remain unanswered and products are not adapted to the latest trends. “What’s more, a poor or delayed flow of information also means customer service employees do not have enough information about the concerns and background of the person they’re talking to,” says Carsten Gauch, head of the IT service development department at BGV Badische Versicherungen in Karlsruhe. This leaves a lot of business potential untapped overall.

The task

In the second phase of the innovation network, which was concluded at the end of 2020, Fraunhofer IAO and the insurance companies particularly focused on ways AI can optimize processes in the insurance sector. It could help with tasks from responding quickly to customer inquiries to providing more targeted product recommendations during conversations with customers. “We used the method of process mining in particular as the starting point for our analysis,” says Dr. Kintz.

While the term data mining the automatic analysis of a huge quantity of data for specific characteristics – is already fairly well known among the general public, many people are not as familiar with process mining. “Process mining is the analysis of event logs, which contain timestamps for activities,” explains Dr. Kintz. “These allow you to reconstruct the workflows in an organization: receipt of the e-mail, forwarding, processing, conclusion.”

AI can help bring about positive changes in this area. For example, it can help distribute the workload in such a way that overloading, and the resulting inefficiency, are avoided. In addition, AI-driven assistance systems can analyze texts for key terms to identify customer concerns and automatically initiate steps such as terminating a contract.

The solution

“Among other things, we programmed and tried out a demo version of some AI-driven assistance software in the second phase of InnoNetz,” Pohl reports. This is led to the creation of eLisA, a tool that listens to conversations with customers in call centers and stores information regarding the requests and concerns that are brought to the insurance company. Based on this, the call center employees see real-time recommendations for improved cross-selling on their monitor, i.e. additional products that should be recommended to a caller in a specific situation. At Fraunhofer IAO, AI solutions for the insurance sector have also been developed in projects outside the innovation network and are already in use at the institute: these include the text recognition system Thorpedo, for example, and ARPOS, a checking tool for calculations in motor vehicle damage claims.

“Due to the successful work in the second phase of the network project, together with Fraunhofer IAO, we have moved into the next, more in-depth phase,” says IT expert Gauch from the participating insurance company BGV. However, according to participants on all sides of the network, it will take a great deal of persuasion within the companies for AI-driven support solutions to be accepted at all levels. Until then, there are still many spaghetti knots to be untangled.

What’s important is the ability to make decisions transparent.

Carsten Gauch, Abteilungsleiter IT-Service Development beim BGV Badische Versicherungen, Karlsruhe

“Artificial intelligence isn’t some kind of digital sorcery that makes humans dispensable. For years, our discussions in the innovation network with other insurers and Fraunhofer IAO have shown that AI can improve efficiency and performance in the insurance sector. And that actually helps the human experts in customer service and claims processing. Using digital assistants means they receive better and more up-to-date information, and can make decisions faster and, in principle, more objectively. In the network, process mining gave us deep insights into our own daily workflows, and these were especially valuable to me. They helped us  detect and mitigate vulnerabilities and bottlenecks. It’s important to be mindful of the ethical aspects of using artificial intelligence in the insurance sector as well. We are discussing this issue with Fraunhofer IAO and external experts as well. What’s most important for our customers is that all AI-driven decisions can be made transparent, and at all times. Customers must not get the feeling that there’s a black box working on the other end, making decisions based on unclear criteria.”

Hack with Fraunhofer IAO: Innovate Insurance

In February 2018, as part of the “Innovationsnetzwerk Digitalisierung für Versicherungen” (Digitalization for Insurance Companies Innovation Network), a hackathon took place with 22 interdisciplinary participants from insurance companies.

The goal was to use two basic technologies to create prototypes for chatbots that could answer customers’ questions regarding annual premium statements.