The promise of edge AI in manufacturing is accelerating. From near real-time quality inspection to predictive maintenance that prevents costly downtime, the technology exists to fundamentally transform how factories operate. Yet most manufacturers find themselves stuck in the edge AI computing paradox: they can see the destination clearly, but cannot get there. Pilots stall and data sits in silos. Use cases that work brilliantly in one plant prove impossible to replicate in the next. The ambition and investment are there, yet the gap between vision and execution refuses to close. Tony Judd, Managing Director UK&I, Verizon Business, explains further.
So what is holding manufacturers back? In the conversations I have with C-suite leaders across the UK manufacturing sector, the answer is consistent and not what most boardrooms expect. It’s not talent, budget or technology maturity. It is often something far more fundamental, and far more often overlooked.
The hidden barrier to smart manufacturing
The pressure to close that gap has rarely been more acute. UK manufacturing output contracted in March for the first time in six months, input costs rose at their sharpest pace since October 2022, and supplier delivery times deteriorated at the steepest rate since mid-2022. Manufacturers are leaning harder into AI and automation to protect margin, however the technology is only as good as the infrastructure it runs on.
IT teams spend their time firefighting local issues rather than enabling strategic initiatives, meaning the best technology talent a manufacturer has is often patching yesterday’s problems instead of building tomorrow’s advantage.
Tony Judd, Managing Director UK&I, Verizon Business
Across almost every multinational manufacturer today, the ambition is the same: ‘One IT’. A single, standardised technology environment across every plant, warehouse and office – the prerequisite for everything from a global ERP to enterprise-wide AI. It is where the paradox is finally explained.
The barrier is often the network. Many global manufacturers have grown through acquisition, inheriting a patchwork of regional networks with different suppliers, protocols and service levels. Data sits in silos. Security policies vary. IT teams spend their time firefighting local issues rather than enabling strategic initiatives, meaning the best technology talent a manufacturer has is often patching yesterday’s problems instead of building tomorrow’s advantage.
UK manufacturers are accelerating their focus on AI, IoT and automation, yet the underlying network – the connective tissue that makes those technologies function – often remains fragmented and underpowered. It is like fitting a high-performance engine into a car with no gearbox: all the power in the world, but no way to put it to work.
Why factory pilots break down
Talk to anyone running an edge AI programme in a factory and you will often hear the same stories. The computer vision system that worked perfectly in the demo environment cannot keep pace with a real production line. The digital twin that goes out of sync every time a sensor drops its connection. These are network problems dressed up as AI problems, and they are the reason so many promising deployments never make it from one plant to the next.
The implications extend beyond the factory floor. When network infrastructure is fragmented, so is the data feeding strategic decisions. Supply chain disruptions can take days to assess because information flows through disconnected systems. Leadership teams end up flying blind on the data that matters most, at the exact moment they need to see clearest.
So, the first question for any manufacturer serious about AI is not which models to deploy or which vendors to back. It is whether the foundation beneath them can carry the weight.
What a global transformation looks like in practice
Verizon recently led a multi-year network transformation for a global food producer operating across dozens of sites worldwide. The organisation had grown through acquisition and found itself with the fragmented landscape described above: disparate regional systems, inconsistent service levels and an IT environment that made a unified global ERP virtually impossible to deliver.
The transformation began not with technology, but with people. We brought together international IT stakeholders who had been operating in isolation for years and ran an intensive co-design process, giving every region a voice in shaping the outcome. That proved instrumental in securing buy-in at scale, across regions with very different operating cultures and priorities. When people help design the solution, they are more likely to champion it rather than resist it.
The first question for any manufacturer serious about AI is not which models to deploy or which vendors to back. It is whether the foundation beneath them can carry the weight.
Tony Judd, Managing Director UK&I, Verizon Business
Rather than investing in owned infrastructure with constant refresh cycles, the organisation adopted a Network as a Service (NaaS) model. A unified SD-WAN backbone replaced the fragmented estate, with common SLAs applied consistently across all 70 sites. A fully managed SASE solution delivered integrated security and consistent policy whether a user is on a factory floor in France or working remotely in the UK. Core network functions were virtualised, reducing hardware costs and letting the business scale bandwidth with demand rather than ahead of it.
The outcome is a single, global secure network designed to be faster, safer and more reliable than the patchwork it replaced – the foundation on which the group’s global ERP and wider ‘One IT’ programme now run.
The three factors that make or break ‘One IT’
Looking back, three things mattered more than any individual technical decision we made, which are three of the things I see separating manufacturers who scale their transformation from those who stall partway through.
The first is cross-functional alignment, perhaps the most underestimated element. Local IT teams that have operated autonomously for years will often resist centralisation unless they see clear benefits. This requires intensive engagement, not a top-down mandate: a workshop-driven process where regional stakeholders shape the design and take ownership of the outcome. It is arguably the single biggest predictor of whether a programme survives its first year.
The second is standardised design. When every site – from a production plant in the UK to a corporate office in France – operates to the same standards, the network stops being a collection of regional cost lines and starts behaving like a strategic asset.
The third is a phased approach. No manufacturer can rip and replace its infrastructure overnight, and the ones who try usually break something important along the way. The most effective transformations begin with the backbone, then layer in security and extend progressively to support more advanced workloads like edge AI. Each phase delivers value that builds confidence for the next.
Where transformation actually begins
With a unified network in place, the manufacturer I described now has the infrastructure to pursue what the industry is racing towards: near real-time production monitoring, cross-site data analytics that inform decisions in hours rather than weeks, and the agility to help respond to supply chain disruptions before they cascade.
That is the point too many manufacturers miss. Edge AI, digital twins, autonomous systems – these are not standalone projects. They are outcomes that depend on the quality of the infrastructure beneath them. Transformation is not a single leap; it is a sequence of them. The manufacturers that will lead this next era are not necessarily those with the biggest AI budgets. They are the ones that had the foresight to get the foundation right first and to treat it as the constant that makes every subsequent leap possible.
For more articles like this, visit our Industrial Data & AI channel.
