UK manufacturers are buying intelligence faster than they are building plants capable of using it. Research published by Make UK in June found that 83% of manufacturers are using AI in HR, finance and admin. But in production, the figure is a mere 11%, and in supply chains, it is only 7%. In quality control, the figure is 6%. Just 2% of the manufacturers quizzed by Make UK say that AI is widely embedded across their operations. It would seem that AI has arrived in UK manufacturing – and has promptly got lost on the way to the shop floor. This is not a story about AI being overhyped. It is a story about a queue being jumped. We did not finish Industry 4.0. We renamed it. The connected factory agenda of a decade ago asked some deeply unfashionable questions. Where does your process data actually come from? What condition is that asset in, and how would you know? Who owns the network between the cell and the ERP layer? What happens when a perfectly serviceable PLC installed in 2004 has no route to anything beyond its panel? Those questions were never answered at scale. They were simply overtaken by a more “exciting”, socially commercialised conversation. The evidence that the base layer is missing is not hard to find. While roughly 70% of manufacturers are investing in digital tools, only 10% are operating fully digital factories. Another fascinating discovery is that while there has been a constant rise in the use of connected devices, UK manufacturers report that they using less than half of the operational data they are already collecting. Nowhere is this more visible than in the installed base of drives and controlgear. A modern variable-speed drive is one of the most under-used instruments in British industry. It already knows motor currents, estimated torque levels, voltages, thermal loading and more. In effect, you have a condition-monitoring system and an energy meter in a box that you have already paid for. Yet, on many sites, drives sit on serial links with nothing reading them, or with comms quietly disabled. We are using diagnostic instruments as switches. We need to retrofit that estate. It may not be glamorous, but it is where the value sits: bringing legacy assets into a coherent data architecture, migrating from ageing industrial protocols to secure and semantically useful ones, moving from OPC DA to OPC UA, segmenting IT properly from OT, and building the asset inventory that the National Cyber Security Centre’s OT guidance now effectively expects as a baseline. Installing AI on an un-instrumented plant is rather like fitting a heat pump to a house with no insulation. The technology is not the problem. The building is. This is precisely the territory that Gambica’s members occupy. Representation on the standards committees that determine how industrial systems interoperate and how they are marked and certified is not the sort of thing that generates headlines, but it is what makes retrofit projects viable rather than bespoke. The same applies to our political representation on behalf of our members on energy efficiency and motordriven systems, where the argument for drives has always rested on measurable outcomes rather than enthusiasm. Somehow, there seems to be an obvious link here – technology not being used to its maximum potential, coupled with manufacturers citing skills shortages as the principal barrier to adoption. The engineer who can commission a drive network and the data specialist who can build a model are rarely the same person, and the gap between them is where most industrial AI projects die quietly. None of this is an argument for scepticism. It is an argument for sequence. AI will benefit the plants that spent the mid-2020s instrumenting assets, cleaning data, securing networks and training people, while everyone else was in a workshop discussing use cases. The question for the board is not whether your business has an AI strategy. It is whether your facility could act on one tomorrow morning. Are you making full use of your data? Most UK manufacturers are adopting AI – but in their offices, rather than on their shopfloors. Nikesh Mistry*, Gambica’s sector head for automation, argues that we should be making more of the data that is already being generated by equipment such as variable-speed drives. “The engineer who can commission a drive network and the data specialist who can build a model are rarely the same person, and the gap between them is where most industrial AI projects die quietly.” www.drivesncontrols.com October 2026 45 * Gambica is the trade association for the automation, control, instrumentation and laboratory technology sectors in the UK. You can get in touch with Nikesh Mistry on 020 7642 8094 or nikesh.mistry@gambica.org.uk, or via the Gambica Web site: www.gambica.org.uk
RkJQdWJsaXNoZXIy MjQ0NzM=