This paper describes the architecture of a hybrid decision making system used to implement adaptive, industrial user interfaces. Model-based reinforcement learning is utilized as a technique to generate interface prototypes optimized for a list of different industrial work processes involving an industrial machine. The model-based reinforcement learning approach is powered using both human-derived behavioral data, as well as established UI-centric models such as Fitt’s Law. Human-centric data was sourced via two usability studies featuring three different stakeholder groups, and involving 31 participants. The generated interface prototypes can be loaded by an industrial machine to fit the current work tasks or inform UX designers about the optimal constellation of UI elements to support worker’s efficiency and behavior.