By Tom Howarth, Manufacturing Specialist. Estimated Reading Time: 6-8 mins Date: 3rd August 2026
In-Depth with GM Business Growth Hub - A series of longer reads from the Hub, designed to give you comprehensive industry insights. More words, more knowledge.
Let’s start with a caveat: I am not an AI expert. However, based on consistent engagement with manufacturers across Greater Manchester, these are grounded observations rather than theory. While there are always exceptions, the patterns are remarkably consistent.
Our team of Manufacturing Advisors works predominantly with small and medium-sized manufacturers rather than large multinationals with deep pockets. These are owner-managed businesses balancing production, finance, sales, and people management, often simultaneously. In many ways, they are the backbone of the regional economy: stable, high-value employers with deep local roots.
Every business has heard of AI. Many have experimented with it. A smaller number have implemented it in some form.
What we are not seeing, however, is the much-promised instant productivity uplift. Instead, a familiar manufacturing reality persists; gains made in one area are often neutralised elsewhere. The system performs only as fast as its slowest constraint.
This is not a criticism of AI. It works, it is here to stay, and those who adopt it effectively will gain a competitive advantage. But the key point is this: AI must be integrated into a system, not deployed as a standalone intervention.
There is nothing new in this principle. Production managers have long understood that installing a machine capable of operating at twice the speed of its predecessor does not double output; it simply increases work in progress (WIP) if downstream constraints remain unchanged. This is a classic Theory of Constraints principle.
Similarly, MRP and ERP systems are often presented as comprehensive solutions to production planning. In reality, many implementations fall short of expectations. Complexity, poor data quality, and misaligned processes frequently result in businesses using only a fraction of available functionality.
Tom Haworth, Business Adviser & Manufacturing Specialist
Start with the Process, the System, and the People
- Define Workflow
- Clarify Roles & Responsibilities
- Apply Big Systems Thinking
- Identify Bottlenecks & Efficiency Issues
- Make Continuous Improvements
The solution is not to start with AI, but to step back and understand the end-to-end manufacturing system; both in terms of material flow and organisational knowledge.
Process mapping remains fundamental. While there are now excellent digital tools available for mapping processes and developing standard operating procedures (SOPs), there is still something uniquely effective about simple methods like brown paper and Post-it notes. They are accessible, inclusive, and make the process visible to everyone—crucially enabling people to challenge assumptions with “it doesn’t actually work like that.”
Effective process mapping allows businesses to define the full end-to-end workflow and clarify roles, responsibilities, ownership, and decision points. Mapping also facilitates broader understanding of how the wider system operates and where constraints exist. Making it possible to identify internal and external bottlenecks and expose any inefficiencies, delays or rework loops.
Importantly, this step alone - without any AI - often delivers measurable productivity improvements.
It is also essential to recognise that optimisation is not just about process, but about the system and the people operating within it. Planning must consider the system as a whole, as well as the role design and operational reality.
In many cases, roles should not be removed but reframed. Ownership and accountability remain with people, while AI and digital tools act as enabling layers or gateway processes that support better decision-making rather than replacing responsibility.
Start Small and Build Momentum
One of the biggest mistakes organisations make is attempting to transform everything at once.
Manufacturers have long understood the following principle: how do you eat an elephant? One bite at a time.
AI adoption is no different - the most successful implementations start small, with a single repetitive task, a clearly defined process, or a known operational constraint. In many cases, this does not even begin with AI. Conventional automation or digitisation often delivers faster, clearer, and more reliable returns.
There is an important distinction here that is often blurred in discussion: automation, digitisation, and AI are not the same thing. Many SME manufacturers are being presented with “AI solutions” when a simpler automation or system improvement would solve the problem more effectively.
Once value is proven in one area, it can then be expanded and integrated into the wider operational system. This staged approach reduces risk and builds organisational confidence.
It is also important to recognise that AI is unlikely to be a single, organisation-wide system. More often, it will consist of multiple tools and models that must integrate with each other, as well as with existing ERP, MRP, and operational systems.
Digitisation vs Automation vs AI Implementation
One of the most common sources of confusion in SME manufacturing is the assumption that digitisation, automation, and AI are interchangeable. In practice, they represent three very different stages of maturity and getting the order wrong often leads to poor returns.
Digitisation
This is the foundation, it is the process of converting analogue information into digital form. This might include moving from paper-based job cards to digital work instructions, replacing whiteboard planning with an electronic scheduling system, or capturing production data in a structured database. Digitisation does not change the process itself; it simply makes it visible, measurable, and accessible. Without this step, any attempt at automation or AI is fundamentally limited by poor or incomplete data.
Automation
Building on digitisation, automation is the execution of defined tasks with minimal human intervention. This can range from simple rule-based systems (for example, automatic reorder points in stock control) through to physical automation such as CNC machining, robotic handling, or automated inspection systems. Automation is best suited to stable, repeatable processes where the rules are clearly understood. It delivers efficiency by removing variation and reducing manual effort, but it still operates within a fixed logic.
AI implementation
Takes this process a step further by introducing systems that can interpret data, recognise patterns, and support or optimise decision-making. Unlike traditional automation, AI can deal with variability and uncertainty; such as forecasting demand, predicting machine failure, or optimising production schedules based on multiple changing constraints. However, AI is heavily dependent on the quality, structure, and consistency of the underlying digitised data and the stability of the processes it is applied to.
A useful way to think about the relationship is this:
- Digitisation makes the process visible
- Automation makes the process repeatable
- AI makes the process adaptable and optimised
Crucially, these are not substitutes for one another. Each layer depends on the strength of the one beneath it.
This is why many SME manufacturers see limited returns when they “jump straight to AI.” In reality, they are often attempting to optimise something that has not yet been properly digitised or standardised.
Tom Haworth
Where AI Fits
At this point, AI becomes relevant as an enabler rather than the starting point.
Manufacturers who have engaged with Lean methodologies, Six Sigma, or structured problem-solving approaches will already recognise where waste exists. Digitisation and AI simply expand the range of tools available to address it.
However, the fundamentals remain unchanged:
- Optimise workflow before automating it
- Reduce inventory before accelerating throughput
- Eliminate defects before scaling production
Introducing AI into a poorly designed process simply allows you to do the wrong things faster.
Practical Considerations for SME Manufacturers
Before implementing AI, manufacturers should ask:
- Is the underlying process fit for purpose?
- Can elements of the process be standardised or simplified?
- Would conventional automation or robotics deliver a clearer return?
- Where can AI genuinely add value (forecasting, quality inspection, scheduling optimisation, predictive maintenance)?
- Are the right people involved in reviewing and improving the system?
- Is there a culture that supports change and continuous improvement?
In addition, there are increasingly important governance and operational considerations:
- Data quality and availability; AI outputs are only as good as the data provided
- Cyber security, intellectual property, and data sovereignty; businesses should understand where their data is coming from, where it is stored, and how it is used, particularly when using global providers
- Integration with existing systems (ERP/MRP); avoiding standalone tools that create further silos
- Workforce capability; upskilling staff to understand and use AI effectively
- Change management; ensuring adoption is sustained, not resisted
Frameworks such as Lean Manufacturing and structured problem solving provide a more reliable foundation for adoption than technology-first approaches.
A Regional Perspective
Tom Haworth
Conclusion
For SME manufacturers, the competitive advantage does not come from adopting AI quickly, but from adopting it well. That means understanding processes, designing systems properly, defining ownership clearly, and applying AI where it genuinely removes friction rather than adding complexity.
The takeaway is simple: Fix the process first. Design the system properly. Then use AI to accelerate what already works.
Those who follow this approach will not only see productivity gains, but will build more resilient, scalable, and competitive manufacturing businesses for the long term.
More About Tom
Tom Haworth is an award-winning Manufacturing Advisor with expertise in manufacturing, innovation, and business growth.
Specialising in food manufacturing, he helps businesses improve productivity, optimise processes, and implement continuous improvement.
With entrepreneurial experience as founder of TechnoSpark, Tom combines strategic insight with practical delivery, supporting manufacturers to embrace innovation, strengthen operations, and achieve sustainable growth.
