In the Spanish industry there is a lot of talk about digital twinsbut very little of what happens when they are injected with real AI and taken out of the individual machine and placed in the context of an entire plant. It’s right there where Tobias Thelemannproduct manager for mechanical components and electrical installation at Reichelt electronicslocates the real qualitative leap: the passage from a twin that predicts isolated breakdowns to another that is capable of understanding chains of complex failures that cross lines, machines and factories.
Tobias Thelemann argues that AI does not simply make the digital twin “smarter”, but rather turns it into a system capable of explaining why a failure occurs, where it originates and what impact it has on the entire production process. The twin stops being a reactive tool that triggers local alarms and becomes a living model of the plant, capable of relating small deviations at multiple points and anticipating chains of events that no human team could systematically analyze.
For industrial CIOs and CTOs, this means moving from putting out fires to managing availability, quality and costs at a granularity previously unknown.
However, the main brake, he warns, is not in the algorithms, but in the data and the organization. The reality of many Spanish factories is that of a legacy machinery park, heterogeneous in manufacturers, interfaces and formats, with very disparate data qualities and systems that were never designed for comprehensive data-based scenarios.
Added to this is the lack of clear roles that assume continuous responsibility for the quality of the data, the models and their daily exploitation, which condemns many digital twin projects to remain an interesting pilot instead of becoming a plant structural tool.
Therefore, the roadmap proposed by Tobias Thelemann is deliberately pragmatic: start small, on a critical machine or a process stage where stops really hurt, organize the data, apply simple models and quickly translate that knowledge into tangible operational decisions, such as fewer unplanned stops or more predictable maintenance.
(MCPRO) For a Spanish industrial CIO/CTO, what is the real qualitative leap that AI and machine learning contribute to the evolution of digital twins? That is, what is the difference between a digital twin that detects an imminent failure and one that anticipates complex failure chains by crossing data from multiple machines?
(Tobias Thelemann) I think the real qualitative leap is not so much that AI makes a digital twin “smarter,” but that it allows it to think in a fundamentally different way. A classic digital twin basically focuses on a single machine. It detects that something is wrong and that this or that failure will probably occur soon. That’s helpful, but it’s still very local and limited to current red flags.
With AI and machine learning, the approach completely changes. The digital twin no longer only recognizes symptoms, but begins to understand the relationships between them. By crossing data from several different machines, production lines or even plants, it identifies patterns that, at this level of complexity, are practically unapproachable by human decision processes or can only be analyzed with considerable effort. Suddenly, it is no longer just a matter of a specific machine being subject to breakdown, but rather how small deviations from different sources can turn into a veritable chain of failures.
The difference, therefore, is that the digital twin as it has been used until now predicts what is likely to fail. The AI-supported twin explains why it happens, where the failure originates and what implications it has for the entire production process. Precisely therein lies the added value for the CIOs and CTOs of the Spanish industry, since it allows them not only to react more quickly, but also to intervene actively before costs skyrocket.
From my point of view, the digital twin thus goes from being a simple maintenance tool to becoming a true decision-making instrument. It not only helps keep machines running, but also manages production as a whole in a more stable, predictable and economically efficient way. And that is, ultimately, the revolution that AI makes possible for the Spanish industry.
(MCPRO) If digital twins connect various machines and production stages, crossing data with other sources of industrial information, what are the specific data integration challenges that you see holding back Spanish factories?
(Tobias Thelemann) From my point of view, the biggest obstacles for Spanish factories are not so much in the AI systems as in the data itself. Many plants and production processes work with machinery parks that have been growing historically, where different manufacturers, interfaces and data formats coexist. Correctly integrating all this diversity is, first of all, a complex and expensive process.
Added to this is the great disparity in data quality. While some machines provide detailed information in real time, others offer only minimal data. For a digital twin that must interconnect various production stages, this presents a considerable challenge.
The situation is further complicated because production data is often only partially interconnected. These systems are generally not designed for end-to-end data-driven scenarios. And finally, organizational issues also come into play, such as responsibility for different data sets.
The technology is available, but the path from multiple isolated data sources to a coherent global vision remains a brake for many Spanish factories.
(MCPRO) What real metrics should a CIO measure to justify investing in AI digital twins? What are the real and achievable KPIs in 12-24 months?
(Tobias Thelemann) In the first 12 to 24 months, a CIO’s focus is undoubtedly on KPIs with a direct economic impact, especially reducing unplanned outages. If downtime decreases or maintenance becomes more predictable, the benefit is immediately apparent.
Equally relevant are improvements in facility availability and maintenance costs, for example through more precise use of spare parts or shorter repair times. It is also realistic to influence quality indicators during this period, such as waste reduction or greater process stability.
In a complementary way, KPIs closer to the IT field come into play, such as the accuracy of forecasts or the degree to which the knowledge generated by the digital twin is systematically applied on a day-to-day basis. If clear improvements are seen in these aspects and, at the same time, costs or risks decrease, the investment can be strongly justified.
(MCPRO) What skills do Spanish industrial teams lack to take real advantage of digital twins?
(Tobias Thelemann) In my experience, the main challenge in the Spanish industry lies not so much in the technology itself, but in the ability to optimally connect production and IT competencies. Industrial companies have enormous knowledge of processes and excellent engineers, but the integration between production flows and IT systems is not yet fully exploited in all cases. We know how the machine works and we know how computer systems work, but thinking about both areas together remains a challenge.
There are experts in specific areas, but there are few roles that permanently assume responsibility for data quality, models and their use, and these responsibilities are often not clearly defined. Without this clarity, the digital twin runs the risk of remaining an interesting project instead of becoming a fully integrated work tool.
But even with well-defined areas of responsibility, the full potential of digital twins can only be achieved if there is also the right mental focus. Digital twins generate new knowledge, but you need to be willing to base decisions on them and question established processes. This is not always easy and requires brave leaders, willing to explore new paths.
Ultimately, the key is not in the technology itself, but in the ability to fluidly interconnect technology, specialized knowledge and decision-making processes. It is precisely there that it is decided whether digital twins deploy their full potential.
(MCPRO) For a CTO/CIO running industrial infrastructure today, what is the minimum viable roadmap to get started now? How should you raise this with your CEO if there is still resistance or lack of budget?
(Tobias Thelemann) In my view, it is essential to deliberately keep the barrier to entry low. For a CTO or CIO, it is not about immediately deploying a grand digital twin strategy, but about starting pragmatically and on a small scale. The first step is to focus on a clearly defined area, for example, a critical machine or a stage of the production process in which failures have truly perceptible consequences. There, data is collected, integrated in an orderly manner, and simple analysis models are used to generate transparency. Even without big promises, but with the aim of obtaining, first of all, reliable knowledge.
The next step is to incorporate this knowledge into concrete decisions, such as more predictable maintenance planning or reducing downtime. When the first positive results begin to appear, the digital twin can be gradually expanded, connecting more machines or new data sources. The important thing is that each step provides a clear benefit and does not become a mere technological project.
To create a realistic expectation, it is key to inform the CEO from the beginning about the limitations. It should not be presented as a purely AI-based vision, but as a learning project with controlled risks and an acceptable budget. Instead of talking about algorithms, it is better to talk about fewer failures, more stable production and rapid learning. Especially when there is resistance, it helps to make it clear that doing nothing is often more expensive than taking a small, controlled first step. When the first use case generates a measurable benefit, the budget for the next steps usually comes almost naturally.
