超越试点：降低数字化使能循环供应链中人工智能规模化风险的博士研究员

Most organisations can demonstrate that AI works in a pilot; far fewer can sustain its value when it is embedded in complex, high-throughput operations. In supply chains, this matters not only commercially but environmentally. When AI-enabled forecasting, inventory and decision-support systems lose reliability or user trust, organisations may revert to inefficient practices, contributing to over-ordering, excess inventory, product waste and unnecessary resource use. This project investigates why AI scaling fails in operational settings and how this failure can be anticipated and mitigated. Its central proposition is that scaling failure is fundamentally a socio technical systems problem, rather than simply an algorithmic one. As AI moves from controlled pilots into live operations, validation gates, review checkpoints and governance workflows operate under continuous operational pressure. Rather than being deliberately removed, these controls may gradually degrade or be bypassed as throughput demands increase, allowing drift, unmonitored errors and declining resource-efficiency benefits to emerge.

Food and perishable goods supply chains provide a particularly valuable empirical setting. Forecasting accuracy, inventory decisions and operational controls have immediate consequences for spoilage, waste, resource consumption and carbon emissions. The South West's substantial agricultural and food economy therefore provides an important regional context in which to investigate how AI can support more productive, resilient and resource-efficient supply chains, while generating insights transferable to other time-critical sectors. The research will comprise three integrated phases. - First, a systematic review will map evidence on AI scaling, governance mechanisms and failure modes in operations and supply-chain contexts. - Second, a multi-case empirical study will examine organisations at different stages of AI deployment, combining interviews across strategic, operational and frontline roles with analysis of available deployment and operational data to identify where and under what conditions governance and validation controls degrade.

Engineering and Physical Sciences Research Council Doctoral Landscape Award - Third, an agent-based simulation will model control degradation under varying levels of operational load. The resulting scaling-readiness framework will then be validated through a survey of operations professionals. The project will deliver both theoretical and practical contributions: a socio-technical model explaining how AI governance and validation controls behave under operational pressure, and a diagnostic instrument to help organisations assess their readiness to scale AI while protecting the resource-efficiency gains that motivated adoption. The project also advances digitally enabled circularity. Circular systems depend on reliable information about material flows, product condition, demand and resource use. Maintaining trustworthy AI-enabled decision support is therefore an important enabler of effective circular operations and value retention.

The student will join a supervisory environment spanning sustainable operations, analytics, AI and computational modelling, with access to the ECCE, INDEX and the IDSAI. An extended industry placement will provide exposure to live operational challenges and knowledge exchange. Training will span computational simulation, mixed-methods research, AI governance and industry engagement. Applicants from engineering, computer science, data science, operations management, information systems and related disciplines are welcome, particularly those with experience of data-driven or AI-enabled operational environments or returning to academic study after industry.

Funded by: EPSRC Engineering and Physical Sciences Research Council £21,805. The studentship will provide funding of fees and a stipend of £21,805 per annum for 2026/7.