Supply chains are piling AI onto foundations that cannot support it

Fragmented systems, poor data and inconsistent processes are leaving logistics groups unprepared to move AI beyond pilots, researchers and industry analysts warn.

Artificial intelligence may be moving fast through logistics and supply chain boardrooms, but many organisations are still trying to run it on foundations that were never designed for automation at scale. That is the central warning from Genpact and HFS Research, which argue that the sector is carrying a heavy burden of “enterprise debt” made up of fragmented systems, regional silos and manual workarounds. In practice, that means companies often have the appetite for AI, but not yet the operating discipline required to make it consistently useful.

The problem is especially sharp now that agentic AI is entering the conversation. These systems are meant to do more than analyse information; they are expected to take action, whether by rerouting freight, flagging a delay before it escalates or managing exceptions with limited human input. But as Genpact notes, that only works when the underlying data is trusted, the workflows are clearly defined and the business rules are consistent. Where shipment visibility, asset tracking and supplier records remain split across different platforms, the promise of AI is weakened before it has even begun.

Research cited by ITPro, based on Dell Technologies’ findings, points to a similar gap in enterprise readiness more broadly. Nearly two-thirds of EMEA IT leaders say their infrastructure is not prepared for production-scale AI and analytics workloads, while more than half say security and compliance worries have already slowed initiatives. In the UK, the picture looks particularly strained, with a large majority of leaders reporting that siloed approaches to AI, workplace technology and modernisation make operations harder to manage. The same research suggests many organisations still lack a unified roadmap for AI, data and security, even as they face pressure to be more selective about spending.

Academic work supports that diagnosis. A review published in Springer emphasises that AI adoption is frequently hindered by poor data quality, limited standardisation and difficult-to-share information across supply chain partners. The authors note that supply chain data is often unstructured, expensive to clean up and hard to generalise across different use cases. A separate review in ScienceDirect similarly identifies multi-actor collaboration, disparate data sources, weak governance and change management as recurring barriers to implementation across supply chain functions.

That helps explain why many AI projects in logistics struggle to move beyond pilots. According to Avenga, some of the most common obstacles arise before a model is even selected: incomplete data, inaccessible ERP or warehouse systems, model drift, security concerns and low trust in outputs that are difficult to explain. The result is often an uneasy compromise, in which teams automate parts of a process without redesigning the process itself. In that situation, AI may accelerate existing weaknesses rather than fix them.

Genpact argues that the more practical response is to work from the bottom up. First comes data unification, so that shipment, asset and supplier information can be reconciled into a single operating view. Then comes the harder task of redesigning exception-heavy workflows, so that routine issues can be handled by systems while humans focus on the most complex judgement calls. Finally, leaders need to be candid about where the biggest drag lies, because process and technology debt often sit at the heart of the problem even when they receive less attention than more visible investments.

The broader point is not that supply chains should hold back from AI. Rather, it is that the value of AI depends on whether the network beneath it can support the workload. If organisations continue to layer new tools on top of old structures, they will keep running into the same bottlenecks, no matter how sophisticated the model becomes. The companies most likely to gain an advantage will be those willing to confront the messy, unglamorous work of rebuilding the system underneath.

Source Reference Map

Inspired by headline at: ^[1]

Sources by paragraph:

Source: Fuse Wire Services