AI in Manufacturing – Different Shop Floor, Similar Challenges?

The “Imagination Gap” is wider in manufacturing than at the desk

One phenomenon that particularly inhibits AI adoption in manufacturing is what researchers call the “Imagination Gap” – the widespread difficulty organisations face in concretely envisioning what AI could do for their specific operations. The term describes not a lack of intelligence, but a structural problem: without reference points and tangible examples, there is simply no foundation for identifying meaningful use cases.

A 2025 study by RBC and the University of Toronto’s Munk School identified this Imagination Gap as the central obstacle: not a shortage of technology or talent, but a profound difficulty in recognising concrete applications.[12] The study describes several sub-phenomena that are directly transferable to manufacturing SMEs: a “Paralysis of Abundance” when confronted with too many theoretical possibilities, a “Decision Paralysis” when selecting a first use case, and “Champion Attrition” when internal advocates leave the organisation and take the momentum with them.

Deloitte partner Jason Bechtel captures the phenomenon vividly: one has, so to speak, a magic lamp and can wish for anything – but people freeze. It suggests, he argues, that the imagination muscle, or indeed the curiosity muscle, has atrophied.[12]

This Imagination Gap is wider in manufacturing than in administrative settings for a straightforward reason: office workers have already had direct contact with AI through ChatGPT, Copilot, and similar tools, gaining first-hand experience of what the technology can do – even if they don’t use these tools systematically. Workers in manufacturing lack this reference point entirely. There is, as yet, no equivalent of ChatGPT for the CNC machine or the assembly line that would make AI’s possibilities tangible and experiential.

The German federal government has recognised this gap. The BMBF-funded ProKI-Netz programme – a demonstration and transfer network for AI in manufacturing – was established in 2022 to show small and medium-sized enterprises what AI can do through physical demonstrators. The BMBF funding phase ran from 2022 to 2024; the network has since been integrated into the structures of the Scientific Society for Production Engineering (WGP) and remains active across eight research centres at leading German universities.[13] Similarly, the Green-AI Hub Mittelstand, funded by Germany’s Federal Ministry for the Environment (BMUV) and the German Research Centre for Artificial Intelligence (DFKI), has now completed more than 20 pilot projects with German SMEs, publishing the results as open-source solutions.[14]

Both programmes exist because businesses need support in identifying and evaluating AI use cases in production – and tangible demonstrations alongside guided pilot projects provide exactly that. Systematic enablement – the targeted empowerment of employees and business leaders to recognise and act on AI potential within their own working context – closes this gap more sustainably than any technology purchase.

European SMEs are proving it works

Despite the well-documented barriers to adoption, a growing number of SMEs are demonstrating that AI in manufacturing delivers concrete results – without requiring multi-million budgets or large-corporation infrastructure.

Heismann Drehtechnik GmbH, based in Velbert, Germany (135 employees), began developing an AI-powered system for defect detection and automated readjustment on CNC multi-spindle lathes in October 2024, working with the DFKI through the Green-AI Hub. Material variations had previously caused production interruptions requiring labour-intensive manual readjustment. The AI system now analyses machine data, detects deviations early, and provides cross-operator recommendations for parameter adjustments. Managing Director Martin Gawenda reports that it was possible to noticeably reduce scrap within a single month. The pilot implementation took six months.[14]

Mühlhoff Umformtechnik GmbH, based in Uedem, Germany (approximately 400 employees, automotive supplier to BMW, Daimler, and VW), developed an AI-based image recognition system for container inventory management in 2023/2024, working with the Fraunhofer IML through the Mittelstand-Digital Centre Ruhr-OWL. Camera-based computer vision identifies and classifies container types in outdoor storage areas automatically – regardless of weather conditions. Unnecessary forklift journeys are eliminated.[15]

A survey of 55 top decision-makers from leading SMEs and family businesses, conducted as part of the “Zukunftsmacher” (Future Makers) study in November 2025, found that 64 per cent reported measurable efficiency gains from AI – some as high as 80 per cent.[16]

The challenge, then, is not whether AI works in manufacturing SMEs. It demonstrably does. The challenge lies in moving businesses beyond the threshold of initial experiments, pilot projects, and wait-and-see attitudes.

What this means for business leaders

The parallels between AI adoption in office environments and in manufacturing are no coincidence. They point to a fundamental insight: AI implementation is first and foremost a leadership and change management challenge – regardless of whether the affected processes take place at a desk or on a production line.

For managing directors of manufacturing SMEs, three priorities emerge:

First: The assumption that AI in manufacturing necessarily requires major investment in robotics and sensor technology is outdated. Many of the most impactful applications – from automated workplace safety analysis to predictive maintenance and production planning – rely on existing data and accessible tools. A joint VDMA/PwC study notes, however, that more than half of the DACH-region companies surveyed are budgeting less than €100,000 for AI investment.[17] That can be sufficient to get started – provided the funds are deployed strategically.

Second: The Imagination Gap cannot be closed by purchasing technology. It requires systematic enablement – the targeted empowerment of teams to recognise and act on AI potential within their own working context. Manufacturing employees need the opportunity to experience AI applications in their environment and to develop their own ideas. The companies identified as successful in the research – Heismann and Mühlhoff – found this path through structured partnerships with research institutions. Not through top-down technology mandates.

Third: Treating AI adoption as an IT project will lead to the same sticking points in manufacturing as in administrative functions. The formula – 70 per cent people and processes, 20 per cent technology, 10 per cent algorithms – applies to the production floor just as it does to the office. Change management, clear process documentation, and visible leadership commitment are not soft factors – they are the hard prerequisites for measurable success.

The Bitkom study summarises the situation succinctly: 42 per cent of German industrial firms say they lack the expertise to integrate AI into their processes.[4] This is not a technology problem. It is a leadership task – and leadership tasks can be resolved: systematically, step by step, and without a big bang.