Systematic testing of multimodal GenAI

AI has arrived in everyday insurance: it classifies documents, answers questions in the form of phone bots, helps to better assess risks and automates non-critical decisions, for example in dark processing. But what has primarily made processes more efficient so far is now becoming a source of new ideas, decision-making processes and content. Our colleague Georgina Neitzel, Head of the ERGO Innovation Lab in Berlin, reveals exactly how in the following guest article, which was first published by IT Fachmagazin.

Multimodal generative AI (GenAI) marks the next stage in the evolution of AI, with possibilities that go far beyond process optimisation. At the ERGO Innovation Lab, an incubator for new ideas and spin-ins within the ERGO Group, we are already exploring the potential of this technology and turning GenAI from an idea into a source of value creation.

This is because the simultaneous processing of language, text, images and videos in real time enables completely new human-machine interactions.

At the ERGO Innovation Lab, ideas are turned into concrete solutions

At the ERGO Innovation Lab, we have already gained extensive experience with GenAI-supported product development. We have not only made classic innovation processes such as design sprints more efficient, but also later phases such as user experience optimisation. GenAI helps us iterate solutions faster, whether through AI-supported evaluation or the efficient bundling of user feedback. This is a decisive factor in bringing products to market faster.

In addition, GenAI will open up completely new avenues of communication in the future. For example, AI-generated videos could help explain complex products in a more understandable way. Virtual avatars could guide customers through simple requests around the clock, such as providing support for damage reports. Internal chatbots could give operational teams faster access to knowledge databases. At the ERGO Innovation Lab, we are testing how this can be implemented in practice. How GenAI can be used to specifically improve service quality while relieving internal teams of routine tasks.

Language models such as ChatGPT or AI summaries from Google make it possible to obtain immediate answers to complex questions and even product recommendations in natural, easily understandable language. We are therefore currently conducting an intensive study to determine how we need to align our digital content in the future in order to remain visible and competitive in this changing search environment.

GenAI is changing the process of risk assessment and acceptance by insurers

Another key area we are focusing on at the ERGO Innovation Lab with regard to GenAI is underwriting, i.e. the process of risk assessment and acceptance by insurers. Here, GenAI could help to efficiently merge diverse data sources, from structured information to unstructured images, text and language, enabling, for example, more individual risk assessment and more accurate quotations.

We have recognised that this technology goes far beyond process optimisation and paves the way for new data-driven business models by comprehensively identifying risks, mitigating them at an early stage and enabling tailor-made policies.

We are also tapping into the field of intelligent automation with Agentic AI, a form of AI that makes autonomous decisions, acts proactively and adapts independently to changing environments or requirements. An example of Agentic AI is a system that independently processes damage reports and obtains missing information. Agentic AI can help structure complex ‘task packages’ and, for example, link information from different sources and provide preparatory analyses. This reduces manual effort and optimises information flows, allowing specialist departments to focus more on interpreting the results and deriving concrete measures. It is important to note that at the ERGO Group, we take a human-centred approach to the development and use of AI-based solutions. As the ‘human in the loop’, people always have the final say.

Even ideas that don't scale can help you move forward

As promising as the above developments in GenAI are, one thing must always be clear:

As part of our innovation work, we systematically test technologies and their use cases from various perspectives, such as strategic fit, technical feasibility, user acceptance and market potential. Some digital projects are discontinued entirely because they do not meet expectations. Recognising early on when hypotheses do not hold up in validation is crucial for responding agilely, learning from the experience, deriving improvements and freeing up capacity for truly promising projects. Every validation provides valuable insights, regardless of whether the hypothesis holds up or not. A systematic validation approach and a culture of learning are therefore essential. And they form the basis for successful innovation management.

Text: Georgina Neitzel, Head of ERGO Innovation Lab