A research team at Graz University of Technology asked itself exactly this question – and went to extraordinary lengths to answer it. Instead of theorizing, they simply put it to the test: They automated the development of a complex product using AI agents they developed themselves. Requirements verification, architectural models, system simulation, parametric design, test cases – all automated. The result: proposed solutions in minutes, where it previously took weeks or even months.
Does this sound like the end of PLM as we know it? Quite the opposite is true.
The real key to success wasn't AI
The team was able to implement the automation only because it had a consistently linked PLM database that had grown over the years: requirements models, architectures, and workflows in Teamcenter, all neatly maintained and interconnected. Without this foundation – according to the team – the project would not have been possible. Notably, the PLM system appeared as a data source in nearly every step performed by the AI agents.
It is also interesting to note where the system reached its limits: As soon as completely new requirements came into play for which there was no existing database, human intervention was still required. Automation, therefore, does not replace engineering – it builds upon it.
The real problem: a lack of connections, not missing data
An analysis of the company’s own digital threads revealed that while PLM data was available, the links between them were often missing. Some of the knowledge “resided” only in the minds of individual employees – not in the system. The actual research finding was thus not a tool-related problem, but rather a connectivity issue.
This sheds new light on a debate that is currently occupying many PLM decision-makers: the idea of completely replacing traditional PLM with a knowledge graph. The presentation offers a nuanced perspective on this – today, PLM systems are in fact mostly just a network of linked objects, not a semantic knowledge graph that also maps the meaning of these relationships. However, knowledge graphs are not a surefire solution either. Their construction is still largely manual, and automated linking suggestions continue to require manual refinement by subject matter experts. The cause of incomplete data thus rarely lies with the tool itself – but rather, in most cases, with a lack of modeling guidelines and their consistent adherence within the company. This is also a management responsibility.
An Expensive Wake-Up Call from the Real World
A case mentioned in the presentation illustrates just how serious this issue can become: A German company faced a claim for damages in the tens of millions before the High Court of Justice in London – because it was no longer possible to reconstruct the methods and data used in earlier calculations and tests. The company ultimately won the case, but only through a labor-intensive and costly reconstruction involving new measurements and tests.
The result: Workflows must be consistently linked to methods. No data record should exist in the system without a traceable link to a method – not only for the AI automation of tomorrow, but also for governance, compliance, and audit trail requirements today.
The answer: PLM is here to stay – but it’s no longer the only option
The presentation culminates in a three-tiered vision: PLM as a System of Record (supplemented by model-based systems engineering), a knowledge graph as a System of Context that reveals meaning and relationships, and AI as a System of Reasoning that uses this foundation to perform analyses, provide recommendations, and enable automation.
PLM thus has – having “served its purpose well” – a long future ahead of it. However, it no longer stands alone; rather, it is becoming the foundation upon which knowledge graphs and AI can truly take effect.
We've summarized what this means for your company, the specific steps you can take as a result, and how to make your PLM landscape AI-ready in a concise handout:
Would you like to know where your company stands on this issue? Contact us – we’ll work with you to analyze your PLM landscape and show you how to make it future-proof.