Högskolan i Skövde

Discrete manufacturing is subject to challenges posed by changeovers due to diminishing batch sizes and the need for customisation. Concurrently retiring personnel deepens knowledge gaps. Despite the evident potential of AI demonstrated in various studies, the question of scalable implementation for changeovers remains largely unexplored. The present study examines the AI readiness for changeover optimisation at a global bearing manufacturer using pull-based production channel systems. The key challenges identified in this study include complex many-to-many relationships between operations and channels, products re- entering flows, and subcontracting arrangements that affect changeover efficiency. The investigation is guided by two research questions: (1) What AI capabilities provide the highest impact on changeover performance in discrete manufacturing? (2) What organisational readiness factors are necessary for successful AI solution lifecycle management? Using a literature review and case study methodology to examine model channels and changeover procedures, the study reveals significant discrepancies between AI's theoretical potential and practical realities. This work establishes a link between theoretical AI capabilities and practical implementation challenges, thus providing evidence-based guidance for firms evaluating AI opportunities. The key findings highlight that the success of the AI lifecycle depends on organisational readiness. We have also identified the operational AI capabilities required to optimise changeover performance. These offer a foundation for developing frameworks that enable manufacturers to navigate AI implementation while maintaining operational efficiency and leveraging lean manufacturing principles, based on the identified ML selection criteria and organisational readiness factors essential for successful AI adoption.