How AI Is Transforming Engineering in the Packaging Machinery Industry

In March 2026, ChatGPT recorded one billion users per week and one billion queries per day; in 2025, it was the most downloaded app worldwide. In addition, more than 90 per cent of Fortune 500 companies now use ChatGPT Enterprise. AI is therefore far more than a hype. With these figures, Dr Daniel Eckertz, Head of the Innovation Engineering Group at Fraunhofer IEM, demonstrated where AI is heading in his presentation at the Packaging Machinery Conference 2026.

Artificial intelligence has long since arrived in engineering and the packaging machinery industry. “At present, AI is primarily changing the way engineering is carried out, rather than the actual machine design,” says Eckertz. Engineers are using AI, for example, to find information more quickly, create documents, develop software or reuse existing solutions. Particularly in the packaging machinery industry, with its numerous customer-specific variants, faster access to existing knowledge can be a decisive advantage.

From Software 1.0 to Software 3.0

Large language models (LLMs), which Eckertz describes as part of “Software 3.0”, mark the transition from rule-based, deterministic code (Software 1.0) and trained models for recognising patterns in operational and machine data (Software 2.0) towards much more flexible systems. Modern models no longer have to be programmed for every individual task. Instead, they understand natural language and can handle a broad range of tasks.

“In the short term, I see the greatest benefit wherever large amounts of text-based information are processed. For example, in requirements, planning, documentation or the preparation of tests,” says Eckertz. AI tools are also already highly advanced in software development and can deliver significant productivity gains. Design and simulation remain more challenging because AI has to deal with geometries, physical relationships and precise boundary conditions. “In the long term, however, this could also be where a particularly significant lever lies,” the expert emphasises.

Looking ahead, AI could perform design tasks based on natural language. Such applications are not yet suitable for every task. However, they indicate where CAD and engineering software could be heading: away from being purely tools operated by engineers and towards systems that take on specific development tasks based on a description.

Between Technological Potential and Industrial Practice

Technological capabilities are currently developing faster than their implementation in companies. Many mechanical engineering companies are still at the stage of individual pilot applications and assistance solutions.

“Technologically, much more is already possible today than is actually being used in industrial engineering,” Eckertz emphasises. He sees the biggest gap less in the AI models themselves than in their integration into real engineering processes and existing tool landscapes. In addition, data is often distributed, structured differently or insufficiently interconnected. “For me, the next major step is therefore not necessarily an even better AI model, but its seamless integration into everyday engineering,” says the AI expert.

Engineering Knowledge Base as an AI Resource

Mechanical engineering has an extensive body of knowledge that has grown over decades. Drawings, bills of materials, documentation, requirements, software, test reports and previous projects contain information that can be valuable for new development tasks. However, a significant proportion of this knowledge is difficult to access. This is precisely where Eckertz sees one of the most promising applications for AI.

For example, it can link drawings, documentation, requirements, bills of materials and previous projects and make them available in a context-specific way, Eckertz explains. For the packaging machinery industry, this means, among other things, that engineers can access solutions and experience from previous projects more quickly when developing new machine variants.

More difficult is the practical knowledge that is currently held in employees’ heads. This knowledge first needs to be captured and structured. AI can help with this by systematically recording and processing knowledge through dialogue with experienced employees.

Looking ahead, sensors, assistance systems and integrated workplace software could also help derive practical knowledge more extensively from real-world workflows. However, this creates new requirements in terms of data protection, transparency and acceptance. “Companies need to define very clearly what data is collected, what it is used for and where the limits are,” Eckertz emphasises.

However, creating reliable AI applications from this knowledge requires more than simply providing as much data as possible. According to Eckertz, quality, context and the relationships between the information are crucial. An AI system must, for example, be able to understand which requirement relates to which solution, which test and, later, potentially which problem in operation.

People Still Decide

The more deeply AI is integrated into engineering processes, the more important the question of responsibility becomes. Particularly in areas such as machine safety, risk assessment and functional safety, decisions cannot simply be delegated to an AI system. “AI can collate information, point out potential risks or prepare suggestions. The assessment of whether a machine is safe and actually meets the requirements must not simply be delegated to an AI,” says Eckertz.

This becomes even more important because generative AI can produce results that initially appear plausible but may nevertheless be incorrect. “That is precisely why the engineer’s ability to assess the results from a technical perspective will become even more important,” Eckertz is convinced.

The human therefore remains the central authority: they orchestrate processes, check results and retain decision-making authority.

AI is also Changing Business Models

However, AI is not only changing processes and roles within development. The further the technology advances into machine operations, the greater the potential also becomes in the field of service provision.

If manufacturers make their machine and process knowledge digitally available and analyse operational data more extensively, this opens up new opportunities for commissioning, maintenance and service. “I believe that AI will initially make existing services more efficient above all,” says Eckertz. Engineering, commissioning and service could be carried out more quickly and with a stronger knowledge base. At the same time, the technology could give rise to new services: manufacturers could make their machine and process knowledge continuously available to customers through AI-based services or use operational data more systematically for optimisation and maintenance. This could enable packaging machinery manufacturers to evolve their business models from the sale of machines alone towards continuous digital services throughout the entire machine lifecycle.