HDI CEO outlines AI-native transformation and security challenges

Daniel Schulze-Lammers, CEO of HDI-Versicherung, argues that insurance companies must restructure around AI capabilities rather than optimizing existing legacy processes. He identifies two critical architectural shifts: building a 'context layer' that aggregates unstructured data, such as police reports and customer emails, into accessible formats like JSON and Markdown, and redefining workflows around specific decision points rather than linear steps. HDI is currently implementing this strategy in claims processing and corporate customer business, separating deterministic rule blocks from probabilistic AI agent blocks. Schulze-Lammers emphasized that while AI can rapidly analyze thousands of documents to assess risk, human oversight remains mandatory for final decisions due to regulatory constraints. He noted that the industry cannot afford the five-year adoption timeline previously seen for Agile methodologies and must adapt development practices to AI's speed.

To illustrate the evolving security landscape, Schulze-Lammers coded a 1,000-line arcade game called 'Galaxy Invaders' for an internal hackathon with his team. The exercise evolved into a security competition where employees attempted to hack the high-score API. As Schulze-Lammers increased security measures, attackers began using bots to mimic human mouse movements to bypass detection. Approximately five to six employees are currently involved in this ongoing attack-defense simulation. Schulze-Lammers uses this experiment to highlight the rapid evolution of AI capabilities and the need for employees to understand software deployment and security risks in an AI-driven environment, noting that many staff learned IT before AI existed.