Artificial Intelligence in Operations and Supply Chain Management

This thesis investigates how artificial intelligence (AI) can be designed, implemented, and embedded within the operations and supply chain management functions of manufacturing firms, and how these organizations must adapt to absorb the resulting changes. Although artificial intelligence is widely promoted as a transformative technology for manufacturing, the path from a promising algorithm to a trusted, adopted, and value-generating system remains poorly understood, particularly in real industrial settings and small and medium-sized enterprises. Adopting organizational information processing theory as the unifying theoretical lens and combining different interventionist methodologies, the thesis brings together a collection of articles addressing four complementary application domains: automated quality control, artificial intelligence enhanced demand forecasting, interpretable decision-making through intelligent digital twins, and the emerging frontier of agentic artificial intelligence, where autonomous and semi-autonomous agents begin to reconfigure organizational decision-making. Across these settings, developed in collaboration with manufacturing companies in northern Italy, the research traces how AI implementation reshapes information-processing requirements and capacities, surfaces organizational barriers and enablers, and ultimately requires organizations to evolve their structures, routines, and trust relationships. The overarching argument is that the value of artificial intelligence in operations is as much an organizational achievement as it is a technical one, since predictive accuracy is necessary but never sufficient to generate organizational value, which is realized through interpretability, trust, iterative implementation, and the reconfiguration of roles, routines, and relationships. Taken together, the studies connect a closing logic, in which noisy signals and tacit expertise are converted into stable, interpretable, and trusted artifacts, with an opening logic that externalizes organizational knowledge and, at the agentic frontier, progressively delegates decision rights to AI systems. Across this trajectory, the recurring finding is that organizational readiness, rather than model capability, represents the primary constraint on value creation. In low-maturity settings, artificial intelligence’s first contribution is epistemic rather than predictive, as it reveals the state of the organization’s information base before it improves predictive performance, exposing the quality, structure, and accessibility of the knowledge on which subsequent artificial intelligence capabilities depend. The contributions are both theoretical and practical. Theoretically, the thesis extends organizational information processing theory by showing that the fit between information-processing requirements and organizational capabilities is dynamic rather than a one-off structural match, by proposing, from the demand-forecasting study, a perceptual and relational layer through the constructs of data perception, nervousness, and data unreliability, and by recognizing artificial intelligence as both a source of information-processing capacity and a generator of new processing requirements. Practically, it offers interpretable and human-centric pathways through which firms, including resource-constrained small and medium-sized enterprises, can adopt AI in operations, while, at the agentic frontier, proposing a graduated and governed approach to decision autonomy in which segmentation, guardrails, and accountability ensure that increasing autonomy remains transparent and human-centred.