Opportunities For AI In The Insurance Industry

Dr. Sebastian Kaiser is a statistician who enjoys baseball partly because it contains so much data. His interest in language also extends to Klingon: a Star Trek alarm clock once woke him with a sentence each morning. He is comfortable with technical curiosity. What frustrates him is expecting AI to perform magic while ignoring the work needed to make it useful.

In this Innovation Rockstars conversation, ERGO's Head of Machine Learning describes a more grounded approach. Data engineering, business understanding, IT resources and digital processes all matter alongside the model. The team must be able to explain the business case and put the result into operation.

His purpose is also more human than the usual automation narrative suggests. By reducing repetitive work, he wants employees to have more freedom to focus on customers and make those interactions more personal.

Build the surrounding capability, not just the model

Kaiser describes an advanced analytics unit with several complementary capabilities: data engineering, data science, AI engagement managers and dedicated IT support. A production cloud environment provides a place for the work to run.

That arrangement helped persuade him to join ERGO. The team had its own budget and a responsibility to create business value, rather than simply demonstrate what a technology could do.

“But the use case in the end has to pay off.”

The funding gives the team something to contribute when proposing a use case. It does not remove the need to examine the costs and expected benefits. Kaiser describes treating AI work as a normal project within the company, with a business case, even while keeping longer-term possibilities in mind.

For an R&D leader, that changes the staffing question. Hiring strong model builders is important, but the surrounding work needs owners too. Someone has to provide the data, connect the project to an operational need and support deployment within the existing technology environment.

Give translation between teams a real role

The AI engagement managers are a distinctive part of Kaiser's account. Their job spans the analytics team, business units and IT, helping people with different expertise work on the same problem.

They look for use cases, help resolve disagreements and contribute to planning. They need enough understanding of machine learning, insurance and IT to connect the conversations, without having to specialize in every area or write the models themselves.

Kaiser also values the mix of people in the unit. Alongside external hires, colleagues with long experience at ERGO had moved into the team and received training in the new topics. Their knowledge of the organization remained useful as the work changed.

He describes leadership familiar with the company and its ways of operating, rather than an isolated group built around a technical celebrity. The unit does something new, but it still works as part of ERGO.

This helps make adoption more than an explanation delivered at the end. Business knowledge and organizational familiarity are present while the project is being shaped.

Find the pain point before choosing the newest method

Kaiser's team looks for places where automation or optimization can make an operational difference. The work does not have to produce a spectacular demonstration to justify attention.

Staying close to technological developments matters, but he distinguishes that from putting the newest method directly into a running insurance system. The team needs to determine what fits the problem and the environment, then take it through development.

“We are just using a different tool.”

That is part of how he explains the work to colleagues. The objective is not to set up a mysterious technical activity apart from the business. It is to help existing teams address their work with another capability.

He makes a similar distinction between innovation and AI. An insurance product can be innovative without using AI, while machine learning can improve a familiar process without creating an entirely new business. The terms do not have to travel together for the work to be worthwhile.

Connect the model to the customer's experience

Kaiser uses response time as an example of where machine learning can help. A message can be routed to the appropriate person more quickly, and a suggested reply can be reviewed by an employee before being sent.

The model is only part of that workflow. The surrounding process has to carry the message, present the information and make it possible for someone to act. Adding AI without that operational connection would not produce the same result.

He discusses smaller customer-service tasks and out-of-hours contact as further opportunities, while emphasizing the need for digital processes. His broader ambition is to remove repetitive work so people can spend more attention on interactions that need them.

The episode describes these applications as part of his experience and priorities at the time. It does not provide measured savings or response-time reductions. The useful lesson is the connection between a technical intervention and a customer-facing problem, rather than an unsupported performance claim.

Use established expertise to build something different

Kaiser's career highlight comes from his earlier work at Munich Re. He describes an industrial offering where customers paid for output rather than purchasing the machine itself.

What made the example meaningful to him was the combination of capabilities: connected equipment and data, digitization, financial and insurance knowledge, with AI and machine learning contributing to the arrangement. The technology helped make a different commercial model possible.

That story captures the approach running through the conversation. New methods become valuable when they connect with knowledge the business already has and a problem it understands. The organizational work may be less glamorous than the model, but it is what gives the model somewhere useful to operate.