PP Control & Automation calls for a more practical approach to AI adoption in manufacturing

Manufacturers aren’t short on ambition when it comes to Artificial Intelligence (AI), but many are still searching for direction.

Ian Knight, Chief Information Officer at PP Control & Automation (PP C&A), challenges the prevailing, often vague narrative around AI adoption and reframes the conversation around a more practical starting point: operational constraints.

Drawing on real-world application from the shopfloor, he explores how a bottleneck-first approach is helping turn AI from an abstract concept into a targeted, value-driving capability – and what other manufacturers can learn from it.

There is no shortage of intent when it comes to artificial intelligence in manufacturing. The numbers are clear, and they are consistent across multiple studies. Around 80% of manufacturers are planning to allocate a significant portion of improvement budgets to smart manufacturing (Deloitte), whilst over 90% say they are already investing in (or plan to invest in) AI technologies within the next five years (Rockwell Automation).

More than half are already using AI in some form, and four in five believe it will be essential to maintaining or growing their business by the end of the decade (National Association of Manufacturers / Manufacturing Leadership Council).

On the surface, it paints a picture of an industry moving decisively toward an AI-enabled future. However, a closer look reveals a somewhat more nuanced reality. The same body of research highlights that 65% of manufacturers lack the right data for AI applications, and over 60% say their data is unstructured or difficult to use.

Confidence is also an issue. Over 50% of organisations report low confidence in their frontline teams’ readiness to lead AI-driven change, and only a small minority have fully integrated AI into their operational strategy (PwC / Manufacturing Institute). In other words, the direction of travel may be clear(ish), but the route to get there is anything but.

The problem with the current conversation

Against this backdrop, the volume of AI content aimed at manufacturers has exploded. Articles, reports, webinars and conference panels all point to the same overarching message: AI is transformative, adoption is accelerating, and those who fail to act risk falling behind.

Not to say that any of this is incorrect, but much of the advice shares the same limitation. It is broad, high-level, and often detached from the day-to-day realities of manufacturing operations. A typical narrative will focus on themes such as:

• “AI is moving from experimentation to production” (Automation World)
• “Manufacturers must modernise their data foundations” (Deloitte)
• “Integration of IT and OT is critical” (The Manufacturer)
• “AI will reshape engineering productivity and decision-making” (Manufacturing Dive)

Again, all valid points, but for a production leader, operations director or engineering manager, these insights often raise a more immediate question – what does this actually mean for my factory, my processes, and my constraints… right now?!

This is where the disconnect emerges. Because whilst the industry conversation often starts with technologies, platforms, and future-state visions, most manufacturers are still grappling with very practical challenges.

Think bottlenecks limiting throughput, manual processes slowing down engineering or production, skills gaps and knowledge kept inside minds rather than systemised (that’s a big and common one). Think about things like change management through constant engineering updates, and another big one – data that exists but is fragmented, inconsistent, or unusable.

These are some of those operational realities we’re dealing with as a community. And yet, many AI discussions fail to even start here.

Investment is outpacing clarity

The result can be viewed as a growing tension or confusion within the sector. Manufacturers are being told (correctly) that AI represents a major opportunity. They are allocating budget, exploring platforms, and engaging with vendors. But without a clear framework for where AI should be applied, many are left navigating a complex and often overwhelming landscape.

They’re asking where to invest. Should they invest in generative AI tools, predictive maintenance platforms, or quality inspection? Perhaps engineering copilots or digital twins are the correct introduction to AI adoption. The options are layered and expanding quickly. No wonder there’s confusion.

And worth mentioning at this point, that each one comes with its own promise of transformation. However, without a clear link to a defined operational strategy and, importantly, an operational need, these investments risk becoming solutions in search of a problem.

This is why, despite high levels of investment intent, many organisations still struggle to move beyond pilot projects or isolated use cases. AI remains an initiative, rather than becoming embedded into how the business actually operates.

The pace of AI change creates another challenge

There is another factor manufacturers should consider when developing their AI strategy, and that is speed. Manufacturing is accustomed to planning major technology investments over long time horizons. ERP programmes, MES deployments, factory automation projects and digital transformation initiatives are often measured in years.

It is fair to say that AI is different. The capabilities of modern AI systems are advancing at a pace that few industries have experienced before, with new models, tools and techniques emerging every few months and often delivering significant improvements in capability, accuracy, speed and cost.

This creates a new strategic risk. Organisations can spend substantial time evaluating platforms, building business cases, integrating systems and implementing solutions, only to find that the technology landscape has evolved significantly before the project reaches full maturity.

The challenge, therefore, is that long implementation cycles can reduce an organisation’s ability to benefit from rapidly improving capabilities, more so than the challenge of investments becoming obsolete. This is another reason why manufacturers should start with operational constraints rather than specific AI technologies.

A production bottleneck, slow engineering process, fragmented knowledge base or quality issue may remain a constraint for years. The AI tools used to address those constraints may evolve several times within that same period.

By focusing on the problem rather than the platform, manufacturers retain flexibility and remain aligned to a consistent operational objective.

A different starting point: the constraint

A common thread running through the most successful examples of AI adoption in manufacturing is that they do not start with AI, they start with a problem. More specifically, they start with a constraint. Where is time being lost, capacity being limited, and quality at risk?

Or where are decisions slowed or repeated unnecessarily, and where is knowledge trapped in individuals rather than maximised in systems? These questions are far more powerful than asking, ‘how do we adopt AI?’ Why? Because they anchor the conversation in operational reality.

Once a constraint is clearly understood, the role of AI becomes easier to define. It is no longer a broad, abstract capability, instead it becomes a targeted intervention designed to remove friction, improve flow, or enhance decision-making in a specific part of the process.

In some cases, AI will be the right solution. In others, it will not. But either way, you’ll have made an important distinction.

Importantly, this approach aligns AI initiatives with measurable business outcomes, whether that is improved throughput, reduced lead times, enhanced quality, or faster engineering cycles.

The urgency around AI is not misplaced. The competitive landscape is changing, and early adopters are already seeing gains in efficiency, productivity and insight. The risk is that in the rush to adopt, manufacturers overlook the fundamentals.

AI does not fix broken processes. It does not compensate for poor data governance. It does not replace the need for a clear operational strategy. In fact, without these foundations, AI can amplify existing inefficiencies rather than resolve them.

This is reflected in the research. The biggest barriers to adoption are structural: data readiness, integration, skills, and organisational alignment (Cisco), which brings us back to the starting point.

A more practical path forward

The manufacturing sector does not need more reassurance that AI is important, because, well, we get it! That case has been made and there’s no escaping the hype. What it needs is a more grounded, practical way to translate that importance into action.

So, to double down on this idea and to be absolutely clear, my opinion is that for manufacturers looking to move from intention to impact, the question should be ‘where are we currently constrained, and what is the most effective way to remove that constraint?’

At PP C&A, we found ourselves facing exactly this challenge. Like many manufacturers, we could see the potential of AI. What was less obvious was where it should fit within our operational strategy. Rather than starting with platforms, models or technology roadmaps, we started with a simple question – ‘what is our biggest constraint?’

It is no longer a theoretical debate at PP C&A. It is how we are approaching AI in practice. We ignored the platform and technology noise and started by identifying where our operations were constrained. Once those constraints were clear, we engaged specialist AI expertise with a clearly defined brief.

We asked how AI could solve these very specific operational problems. From this, we developed a four-phase implementation strategy. In phase one, we have focused on one of the most time-intensive and resource-heavy activities in electrical engineering – extracting structured data from unstructured documentation.

Technical PDFs (which can exceed 1600 pages) are now capable of being converted into structured, repeatable outputs. This includes embedded rules, full traceability, and the automatic identification of missing components and mismatches. Previously, this process required significant engineering time and manual effort. Now, those same documents can be processed in hours, not days. And this is just the starting point.

The first phase creates the foundation for what follows, with subsequent phases looking at enhancing the structural data, its validation and integration into business systems. Each of the phases builds on the last, moving from data extraction to decision support, and ultimately to execution.

The result is a 36% recovery in headcount capacity, targeting the 60% of time previously lost to manual parsing and interpretation. More importantly, it removes a key point of friction in the early stages of customer engagement, accelerating the transition from enquiry to executable work.

When fully implemented, this will deliver significant gains in time, productivity, and consistency. But just as importantly, it creates something more strategic…defensible intellectual property. Not an AI tool in isolation, but a system built around real manufacturing constraints, with the potential to be deployed not only internally, but as a capability offered to customers and potentially even the wider market.

This approach avoids the trap many manufacturers face today – investing in capability without a clear path to value. Instead, it ensures that every application of AI is directly linked to a measurable operational outcome. The manufacturing sector does not need more reassurance that AI is important. That case has already been made. What manufacturers need is a practical framework for turning intent into measurable results.

For organisations looking to move from experimentation to impact, the starting question should not be ‘how do we adopt AI’ instead, it should be ‘where are we constrained, and what is the most effective way to remove that constraint?’

When manufacturers answer that question first, the role of AI becomes far easier to define. More importantly, it becomes far easier to justify, implement and scale. The companies that gain the most value from AI will not necessarily be those who invest the most, it will be the ones who understand their constraints the best.

PP Control & Automation, which employs over 200 people at its state-of-the-art facility in the West Midlands, is a strategic outsourcing partner for many of the world’s leading machine builders and OEMs, providing module or assembly-based, part or full machine build production capabilities.