AI in UK Manufacturing

UK manufacturing is under real pressure. Rising energy costs, skills shortages and global supply chain uncertainty are compressing margins from multiple directions. AI is increasingly positioned as part of the answer, and in many cases, it genuinely could be.

But there is a significant gap between what AI can do for manufacturers and what most UK factories are currently capable of achieving.

Where adoption stands right now

How you count adoption changes the picture completely. Make UK’s 2026 research into AI, skills and the future of the sector found that only 2% of UK manufacturers have AI widely embedded across their business, and that 43% remain stuck in pilots and experiments. That is not a contradiction of the figures above. It is the same gap measured from the other side. Trying AI somewhere in the business has become common. Running the business on it is still rare.

The same research shows where that gap sits. AI use is concentrated in back-office functions, with 83% of manufacturers using it in HR, finance and administration, against 11% in production, 7% in supply chain and 6% in quality control. The pattern follows the data: back-office systems are structured and connected, so a model has something to work with. Production, planning and quality data frequently is not.

Where the blockages appear

The UK government’s Technology Adoption Review found that only 8% of UK manufacturers had successfully introduced AI and machine learning into their businesses by 2024. When asked why adoption remained low, the most common answer was not cost or the availability of technology. It was knowledge and data readiness.

Predictive maintenance AI, the most commonly pursued use case in UK manufacturing, works by learning patterns in sensor data. It picks up subtle changes in vibration, temperature or pressure that typically precede equipment failure. But this only works when the sensor data is consistent, complete and available in a format the AI can actually use.

Many manufacturers have sensors installed. The challenge is that data is stored locally, collected at irregular intervals or simply not connected to the systems that AI needs to read from. Integration becomes the bulk of the work, not the AI itself.

Quality control AI faces similar barriers. AI can learn to identify defects from images, but only when there are enough labelled examples to train on and when imaging conditions are consistent. Many manufacturers discover that their existing quality records were never designed with AI training in mind.

Supply chain analytics require connecting ERP data with supplier feeds, logistics tracking and demand signals. For many UK manufacturers, these systems were built independently over the years and have never been integrated. That integration work can easily take longer and cost more than the AI deployment itself.

The people dimension

Data is not the only gap. Manufacturing AI also requires people who understand what the AI is doing and can work alongside it confidently.

Maintenance engineers need to understand when to act on a predictive alert and when to question it. Quality teams need to know what the AI can detect and what falls outside its capability. Operations managers need to interpret AI forecasts and factor them into decisions that account for things the model cannot see.

None of this requires deep technical expertise. But it does require clear communication, proper training and a culture that sees AI as a tool to support judgment rather than replace it. Without this, even technically successful AI projects struggle to gain traction with the people who are supposed to use them.

What the manufacturers making progress have in common

The manufacturers moving from pilot to production have generally done two things differently.

First, they started with a specific, well-defined problem rather than a broad ambition to “implement AI.” Not improve maintenance in general, but “reduce unplanned downtime on Line 3 by improving early detection of bearing failures.” A specific problem leads to a specific data requirement, which leads to a realistic plan.

Second, they assessed their data and systems before selecting technology. They knew what they had, what they needed and what it would take to close the gap. That honest assessment prevented months of expensive rework when reality did not match assumptions.

Where to start if you are planning AI in manufacturing

For manufacturers whose pressure sits in cost, supply or service rather than on the asset base, the more accessible entry point is operational data. Demand forecasting, supplier risk scoring, inventory optimisation and quality trend analysis all draw on information that already exists somewhere in the business, in purchasing, production, stock and finance. The obstacle is rarely that the data is missing. It is that it sits in systems that were never designed to talk to each other, so the work of connecting it outweighs the work of applying AI to it.

That is why ERP modernisation and AI readiness are the same conversation more often than they appear to be. A connected operational core is not an AI project, but it is what makes the AI projects that follow achievable rather than experimental. The assessment question is the same in either case: what data do you actually collect, how consistently, where does it live, and what would it take to make it usable?

Your next step

If your AI ambitions sit in planning, procurement, supply chain or finance rather than on the shop floor, the starting question is the same but the foundation is different. Nexer helps manufacturers assess whether their operational data and systems can support the use cases they are considering, and what a practical route to a connected Microsoft Dynamics 365 core would involve. No lengthy consulting engagement. Just an honest view of what is realistic now and what needs to be in place first.

Beyond ERP | The Intelligent manufacturing enterprise

5th November 2026, Microsoft London. Free to attend.

A leadership forum hosted by Nexer and Microsoft for manufacturers working through exactly the questions in this article. A Microsoft keynote, a leaders roundtable, real customer results rather than pilots, and a working session where you prioritise the operational problems AI could realistically solve in your business. Built for business leaders, not technical specialists. Places are limited.

Read more and register>>