The accelerating shift toward industrial AI promises efficiency, agility, and sustainability gains. Still, manufacturers must move beyond the buzzwords and focus on customer value, domain expertise, and a disciplined innovation process to realise its full potential.
The manufacturing sector is no stranger to complexity, yet the volatility of today’s operating environment is pushing even the most traditional industries to rethink how they work. Industrial AI, long an undercurrent of innovation, is stepping firmly into the spotlight. Yet, as Heiko Claussen, Senior Vice President of Artificial Intelligence at AspenTech, argues, success will not be found in chasing technology trends but in a pragmatic, disciplined approach that keeps customer needs at the centre.
Understanding the actual value of industrial AI
The narrative surrounding AI has become increasingly dominated by headlines about large language models and generative systems, but Claussen points out that in the industrial sector, AI has been quietly delivering value for years. “AI has indeed been around for a long time; I have been working in this field for 15 years before joining AspenTech four years ago,” he says. “There has been a wave of interest in AI, especially with the rise of large language models, which has reignited excitement. Many people are only now realising the potential of AI, even though traditional machine learning methods have been widely adopted in industrial applications.
“Often, we do not label these methods as ‘AI’ because we focus on the customer value they bring rather than promoting them as AI for its own sake. AI is simply a tool to solve problems, not an end in itself.” This grounding in real-world problem-solving has become even more critical as executive expectations evolve. Investment appetites are growing, but so is the pressure to deliver meaningful outcomes.
“Executives increasingly see AI as a potential disruptor and part of their role is to prevent disruption from harming their companies,” Claussen continues. “More executives are now driving significant AI investments. We are also seeing new roles, like Chief AI Officer, emphasising the importance of AI at the executive level. However, in fields where safety is paramount, it is essential to implement guardrails around AI technologies, choosing the right tools and ensuring users stay in control.”
Navigating the innovation funnel
The notion that innovation is inherently risky is well understood, yet many industrial organisations still treat AI projects as if they should guarantee a return on investment at every stage. Claussen advocates a more structured approach, using an innovation funnel to manage risk intelligently.
“Research is inherently risky, and traditional development processes do not account for this,” he explains. “An innovation funnel accepts that risk exists in different ideas, so you start with a broad range of possibilities, making smaller investments early on to see how they develop. If an idea does not pan out, it fails quickly and inexpensively.”
This philosophy encourages breadth and pragmatism, ensuring that promising concepts are not lost to inertia or organisational resistance. “Researchers should ideally work on multiple projects simultaneously,” Claussen adds. “High-risk projects might be exciting but may fail. Having a mix of high-risk and lower-risk projects keeps engagement high and allows individuals to maintain a network of contacts with product management, R&D, and customers, which is crucial for the later stages of development.”
Avoiding the common pitfalls of AI deployment
The rush to capitalise on AI’s potential can lead to familiar mistakes. Technology-first thinking remains a persistent trap, as does an over-reliance on data at the expense of domain expertise. For Claussen, starting with customer problems and working backwards is critical.
“The better approach is to identify customer problems first, then find the best technology to address them,” he says. “Another pitfall is relying solely on data-driven approaches. While data can be powerful, domain knowledge is invaluable. Combining data with industry expertise leads to better solutions, and in many cases, this hybrid approach reduces the amount of data required.”
He also stresses the importance of designing intuitive solutions for users rather than requiring specialist knowledge to deploy or operate. “If a solution requires specialised expertise to set up or understand, it might not be effective for the intended users,” Claussen explains. “User experience is vital for adoption, so the solution must fit seamlessly into the customer’s workflows.”
Building the foundations for sustainable AI success
Many manufacturers are eager to embrace AI but are hampered by fragmented or incomplete digital infrastructures. While the dream of Industry 4.0 has been widely discussed, the reality on the ground often lags behind the rhetoric. Claussen believes a pragmatic, hybrid approach offers a path forward even for companies early in their digital journey.
“Even without a complete data infrastructure, we can derive value by understanding the assets,” he says. “In many cases, we cannot gather data from every possible scenario—some set points may be inefficient or even dangerous. Simulation can help fill these gaps, creating models that adapt better to unexpected situations.”
Ontologies, which structure data around asset and sensor relationships, are also crucial for making data genuinely useful for AI applications. Collecting information is insufficient; it must be organised and understood in context to support intelligent decision-making.
“Instead of simply storing data in the cloud, it is crucial to understand how that data relates to operations, failures, and other factors,” Claussen explains. “An ontology-based approach, where data is organised with an understanding of asset and sensor relationships, is essential for making that data useful for AI.”
Embracing agility and preparing for the future
Looking ahead, Claussen sees several forces reshaping the demands on industrial AI. Volatility, demographic shifts, and the drive toward greater efficiency require manufacturers to become more agile and responsive.
“Continuous learning is critical because conditions in the field are always changing,” Claussen explains. “A model trained once might not adapt to changing conditions, so we focus on hybrid models that combine data with first principles.”
Demographic change also presents a challenge. As experienced workers retire and a new generation enters the workforce with different expectations and skill sets, AI systems must automate tasks and actively support faster onboarding and knowledge transfer. “Today’s workforce is accustomed to modern technologies, like large language models, so they expect a similar user experience in industrial applications,” Claussen says.
Efficiency and resilience will also be critical, particularly as companies seek to near-shore operations and reduce their dependency on extended global supply chains. Here again, AI’s ability to automate complex analysis and enable consistent decision-making will be vital.
The road to self-optimising plants
Aspiring towards the self-optimising plant is not a new ambition, but AI now provides the tools to make it achievable. Digital twins, real-time optimisation, and adaptive systems are becoming central to this evolution. “AI enables automation, which is essential for autonomy,” Claussen says. “Continuous tuning is crucial for real-time optimisation, so AI-based digital twins can improve planning and operational efficiency.”
However, regulatory caution will continue to shape the pace and nature of adoption, particularly in safety-critical environments. Balancing innovation with reliability and societal expectations will be essential. “We, as a society, need to find a balanced framework around these capabilities, similar to how other industries approach safety in autonomous systems,” Claussen concludes.
The future of industrial AI is rich with potential, but it will be realised not through technological bravado but through a careful, customer-centric approach that blends data, expertise, and a deep understanding of industrial realities. For executives, the message is clear: focus on building the right foundations, avoid the temptation of quick wins, and commit to innovation that delivers lasting value.
