In high‑precision food manufacturing, even the smallest process deviations can have a noticeable impact on product characteristics and final product weight. This presentation provides insights into the journey of an AI Lighthouse initiative implemented in collaboration with AVEVA and its strategic partner TwinThread. Starting with the identification of a suitable use case, the session explores key questions around relevant measurement data, the extension of existing sensor setups, and the consolidation and contextualization of process data on a shared data platform. Building on this foundation, the presentation outlines how initial AI models were trained, assessed, and evaluated in close connection with process expertise, in preparation towards broader AI adoption.