AI use is quickly becoming the norm in many industries, but the rapid pace at which AI models have developed over the last few years has left many with conflicting feelings.
When you consider that 88% of organisations are now actively using AI in at least one business function, according to figures from McKinsey’s State of AI Report 2025, these misgivings certainly make sense.
However, beneath that startling headline figure lies a more complex reality. Deloitte asserts that only 42% of organisations feel operationally prepared to deploy AI at scale, highlighting a persistent gap between early adoption and meaningful, measurable value.
There is also an important distinction between generative AI, including the large language models like ChatGPT, Gemini and Copilot, and the machine learning models, such as computer vision and digital twinning, for more practical and bespoke applications within manufacturing.
For many manufacturers, the challenge is not whether to invest in AI, but how to ensure that any investment delivers tangible returns. Initial proof-of-concept projects, often focused on predictive maintenance, machine fault diagnosis, or quality control, have demonstrated some potential. However, translating those early wins into consistent, scalable outcomes across the factory floor remains a far more difficult task.
So, how can manufacturers embed AI in a way that supports operational performance, improves decision-making and delivers sustained commercial value?
Workforce dynamics are shifting, with more than 60% of all UK organisations reporting skills shortages. This seems to be disproportionately impacting manufacturers in particular, with 74% asserting that they are being affected by this challenge, according to the Open University Business Barometer 2024.
As experienced operators retire, they can take decades of practical, experience-led knowledge with them. At the same time, manufacturers are dealing with an ever-increasing volume of data. From sensors and maintenance logs to quality systems and supplier inputs, critical insights can often remain buried within fragmented systems or with operators who don’t know how to effectively interpret the sheer excess of data.
Layered on top of this is growing operational complexity; production environments are becoming increasingly interconnected, with multiple systems operating simultaneously across the factory floor. With so many variables competing for attention, this inevitably places a greater cognitive strain on operators, increasing the risk of operational inefficiencies.
Additionally, there are obligations surrounding quality inspection, and traceability has become more stringent in recent years, and there are more requirements for food and allergen labelling under Natasha’s Law.
Taken together, these pressures are sizable, but they are precisely where AI has the scope to deliver value - if it is applied correctly.
The broader debate around AI can feel abstract, but there is already a wealth of practical use cases emerging within manufacturing environments.
One of the most immediate is in knowledge capture and transfer. AI tools are increasingly being used to process unstructured information, such as maintenance records and operator notes, and transform them into accessible, searchable insights. This helps to retain institutional knowledge and makes it available to less experienced staff in real time - creating a user-generated encyclopedia of workplace-specific data.
More than that, AI is also improving decisions on the factory floor. Rather than operators needing to interpret multiple systems and datasets on the fly, AI can cleanly consolidate information and present clear, contextual guidance. In practice, this can significantly reduce the time taken to diagnose faults or respond to issues - and will certainly lessen the administrative burden for operators.
Predictive maintenance is another area where the usable progress of AI is becoming ever more tangible. While the concept itself is not new, AI is making it more intuitive by translating complex data into straightforward, actionable recommendations.
When combined with optical and thermal camera technology on a production line, machine learning can also outperform humans when it comes to quality inspection without slowing down production lines.
Taken in isolation, these applications are not exactly transformational. However, when applied carefully, consistently and at scale, they begin to cultivate meaningful operational improvements which can ultimately result in better financial performance.
Despite rife opportunities, many manufacturers are still struggling to move beyond isolated successes. But reticence in this area is unlikely to be rewarded, as manufacturers that are too slow to implement AI functionally run the very real risk of falling behind their counterparts.
Indeed, in a global context, British businesses lag far behind when it comes to AI and automation: There are twice as many industrial robots per worker in the EU’s manufacturing sector than there are in the UK, and there are four times as many in Germany. South Korea is streets ahead, however, with almost ten times as many robots per worker, with 1,012 robots for every 10,000 workers compared to our 111, according to MakeUK.
Part of the problem here is that AI is often introduced without a clearly defined business objective, which can lead to pilot projects that demonstrate capability but fail to translate results into any measurable action.
Another issue is data, which remains a significant barrier. In many organisations, information is spread across multiple systems, including Enterprise Resource Planning, Manufacturing Execution Systems and operational platforms. Combined, this can make it difficult to create a single, reliable outlook.
Without that foundation, even the most advanced AI tools will only ever deliver limited value, because they will not be presenting the true scope of information - simply optimising another silo to compare and contrast.
The starting point is establishing clarity from the outset. AI should be applied only to clearly defined operational challenges, whether that is aimed at reducing downtime, streamlining decision-making or improving accuracy in inspections. Without a set focus, it becomes difficult to measure success, virtually impossible to justify further investment, and just another process to monitor.
Data integration is equally important in this regard. While this does not necessarily require a complete system overhaul, it does require a more coordinated approach to how information is captured, shared and accessed across an organisation.
Importantly, scaling should be incremental. Rather than attempting to transform entire operations at once, manufacturers are more likely to see success by focusing on a single use case, refining it, and then expanding its application over time. This approach not only reduces risk but also allows organisations to build internal capability and confidence as they go. Over time, this integration is what turns isolated improvements into broader operational gains.
To truly make a dent in improving operational efficiencies, AI implementation must be organically embedded within everyday workflows. More than that, it also requires a shift in mindset: AI should not be viewed as a replacement for human expertise, but instead as a tool that acutely enhances it.
As the market continues to evolve, access to AI itself will not automatically translate into a competitive advantage. Instead, any benefits from its utilisation will come from how effectively organisations structure and apply these tools within their own operations.
For manufacturers willing to take a measured, structured approach, AI offers an ample opportunity to improve efficiencies, reduce costs, radically reduce risk and strengthen long-term resilience. However, realising that value depends less on the technology itself and more on how it is implemented in practice.
Duncan & Toplis provides accounting and business services specifically designed to support manufacturers, including specialist tax and financial planning, R&D support and regulatory guidance.
If you would like to discuss how AI can be used in your business, please speak to your usual Duncan & Toplis adviser or contact Charles Burrell.
