AI—or just software? A buzzword or real progress?

A critical blog post on AI.

“Our solution uses artificial intelligence!”—You see promises like this everywhere these days, from yogurt bottlers to bakery chains to beverage manufacturers. But what’s really behind these claims? Are we talking about self-learning, cognitive systems—or just traditional automation rephrased in new terms? It’s time to set the record straight on one of the most overused terms of our time.

What is true artificial intelligence—and what isn’t?

First of all: there is no clear definition of the term “AI” in an industrial context—and that is precisely where the problem lies. A decision made by software is not AI. Artificial intelligence refers much more to systems that can learn independently based on data, draw conclusions, and solve tasks that would otherwise require human intelligence. Important: AI does not make decisions based on rigid rules, but rather through pattern recognition, probabilities, and continuous adaptation.

Temperature too low, control valve open. What most people today are selling at a premium as innovative and groundbreaking AI is something we’ve been doing for over 25 years. Namely, simple math and logarithms in software that merely create rules. Despite all the added value of automation, this unfortunately has nothing to do with AI.

Thomas BuxGeschäftsführer planemos
A smiling man with a beard and his arms crossed, wearing a light-colored, patterned shirt.

Not AI, but just regular automation:

In industries such as the dairy and Beverage Industry, many processes are now marketed as “AI-powered.” In practice, however, these often involve predefined decision trees (“If the pH drops below 4.3, stop filling”), traditional process automation without any learning logic, or dashboards with recommendations based on fixed formulas rather than adaptive models. As a rule, this is driven by traditional control engineering or standard leveling.

Rule-based control systems (“If A, then B” solutions)

Energy management system for balancing peak loads

Processes with fixed logic or predefined scenarios

Identifying raw material fluctuations and formulation optimization

Visualization tools without a learning feature

Traditional statistical tools

Predictive maintenance

Unfortunately, it is precisely this overuse of AI that lowers expectations and often leads to disappointment when the supposed AI turns out to be nothing more than standard control technology. The result is a loss of trust in technology as a whole, as well as a reluctance to innovate, because the true potential goes unrecognized.

The potential of true AI – where is real intelligence needed in industrial production?

Used correctly, AI can transform industries. Especially when it comes to pattern recognition in the field of big data —that is, the analysis of particularly large datasets—there are numerous useful applications in the automation process. This confirms the realization that without a solid data foundation, an open approach to AI, and a clear understanding of AI, such potential cannot be fully realized. There must always be a willingness to embrace change.

Big data analyses

Fast, responsive pattern recognition for large and numerous datasets, as well as faster and more comprehensive data processing. This makes it possible to quickly identify non-obvious correlations and provide appropriate recommendations.

A person using a tablet with digital checklist and checkmark icons floating above it.

Quality inspection with AI vision

More precise control at higher line speeds, or visual quality control that detects new and previously unknown types of defects.

A modern high-bay warehouse with glowing blue digital data overlays for smart logistics.

Supply chain forecasts

Forecasting fluctuations in raw material supply to improve planning for raw materials, logistics, and personnel.

No AI without the cloud—but there are risks

AI needs data—lots of it, up-to-date, and interconnected. Yet, particularly in the food and dairy industries, there is—and rightly so—a great deal of reluctance toward cloud-based systems. Fearing cybercrime, most companies operate on-premises, meaning they store process data locally. However, cloud solutions are often more secure today than many isolated on-premises systems—provided they are set up correctly.

+ Without centralized data storage, it is impossible to build adaptive AI
+ Without trust in IT security, many companies are hindering their own digital future

Our recommendation

Start small: Don’t put all your data in the cloud; instead, start with non-critical process metrics, for example.

Think about IT security from the very beginning.

Invest in interconnected, scalable data models for the long term to harness the true potential of AI.

Progress without safety is risky, but standing still out of fear is even more dangerous.

Aerial view of the RWTH Aachen SuperC building surrounded by autumn trees.

Digression

Strengthening AI collaboration with universities

Together with the University of Aachen, we have been working for many years on useful AI solutions for our customers. Here’s an example: In a current project, the database identifies the desired modules from the process description by using the recipe logic. So-called Explainable AI (XAI) applications—whose decisions are transparent, traceable, and interpretable for humans—are also very much in vogue.

Conclusion

Education, not hype

AI is a powerful tool when it is understood, honestly described, and used in a targeted manner. It’s time to give the term real meaning again—and to distinguish between artificial intelligence and intelligent software or control technology. Despite all legitimate security concerns, AI offers a real opportunity as soon as large volumes of relevant data are sent to the cloud for analysis.

Let’s talk about AI together.

Whether in production or process optimization, we’re happy to help you identify the true potential of AI—and put it to work for your Company.

Get in touch with us