Kuo-Yi Lin, Shih-Yu Wu, Kotomichi Matsuno
Global supply chains are increasingly exposed to disruptions from pandemics, geopolitical conflicts, and environmental shocks, challenging traditional resilience models. Existing approaches often rely on structured historical data and static assumptions, making them insufficient in contexts with unstructured signals, rapid shifts, and heterogeneous data quality. To address these gaps, this study proposes UNISONE, a Generative AI (GAI)–driven resilience framework integrating disruption sensing, probabilistic state modelling, and adaptive optimisation. The framework has three modules: a GAI component that processes unstructured data into validated signals; a Markov modelling layer that translates signals into probabilistic transitions of capacity, logistics, and policy states; and a Mixed Integer Programming optimisation module that dynamically reallocates resources. This integration enables proactive risk detection, real-time capacity adjustment, and rapid supply chain reconfiguration. Empirical analysis using data from three global regions shows UNISONE sustains service continuity above 90% under disruptions, reduces cost volatility by nearly 20% versus baseline models, and accelerates recovery across cycles. Findings also reveal how data quality and digital readiness shape performance, offering phased adoption insights for industries such as pharmaceuticals, food, and semiconductors. By combining unstructured signal processing, probabilistic modelling, and adaptive optimisation, UNISONE framework advances both theory and practice in supply chain resilience.