This is the second article in my series about the IFIP PLM 2026 conference in Lecce, Italy. Today I want to talk about one of the most anticipated talks of the conference – the PLM Pioneer keynote by Prof. Martin Eigner: “40 Years of PLM”
If you’ve been in this industry for any length of time, you know Martin Eigner. Mechanical engineer, PhD, founder of Eigner + Partner in 1985 (the company whose technology eventually became part of Agile and later Oracle), professor, author, and one of the people who was in the room when the terminology of our industry was being invented.
Forty years in this field gives you a perspective that no analyst report can match and Martin used every minute of it, with his trademark mix of history, architecture, and brutal honesty about what the PLM industry actually delivers versus what it promises.
Who Really Coined “PLM”? And Who Hijacked It?
Martin opened with history, and it’s a history worth retelling. In the mid-90s, together with academic and industry partners, his group coined the term “product lifecycle management” — and the semantic meaning was exactly what the words say: from the very beginning of a product’s life to the end, including recycling. Cradle to grave.
Then, around 1996-97, something happened that shaped the next thirty years: PDM suppliers hijacked the term. Overnight, a PDM supplier became a “PLM supplier” — same system, same functionality, new label. The gap between the original vision of lifecycle management and what vendors actually delivered was baked in from the start, and Martin’s argument is that we have been living with that gap ever since.
His even more provocative point was about why PLM exists at all. In his telling, PLM was created because ERP vendors ignored the engineering data market — they simply didn’t recognize its value. And ALM was later created because PLM struggled to integrate software. His personal conclusion, delivered with a smile but meant seriously: we don’t actually need PLM and ALM as separate categories. What the industry really needs is something like “ERP 4.0” — a unified engineering and business backbone. The current landscape of PLM + ALM + ERP + MES + Excel + shadow IT is an accident of vendor history, not a rational architecture.
I don’t fully agree with the ERP 4.0 framing (single mega-systems come with their own pathologies, more on that in my next article), but the diagnosis of accidental fragmentation is one I share completely.
Three Software Generations Living in One Company
One of the most useful frameworks in the keynote was Martin’s analysis of the three software generations that coexist in today’s PLM landscape:
Generation 1: Monolithic systems. Most of today’s market leaders were created in the 1990s. These are monolithic architectures, now roughly 30 years old, being retrofitted (“weighted by cloud or whatever,” as Martin put it) but still fundamentally monoliths.
Generation 2: Web services / SOA. Starting around 2000 — modular, transparent, open to cloud, easier to containerize. A transitional architecture.
Generation 3: Microservices, cloud-native. The actual state of the art since roughly 2010 — containers, full cloud-native design.
Every industrial company today has to survive with all three generations simultaneously, and that is a big part of why implementation and customization are so painful. Which led to one of the best lines of the conference: “I have never seen a PLM system out of the box.” The marketing says out-of-the-box; the reality is that you customize, then the next version arrives, and your customizations break, and the cycle repeats. Anyone who has lived through a major PLM upgrade was nodding.
Digital Thread: Linked Processes, Not Just Linked Data
Martin’s definition of the digital thread is characteristically precise: the ability to find, select, and analyze connected “information items” across the lifecycle. Based on research (including work presented at this same conference — his co-authored paper with Sven Klaus and Stephan Husung from TU Ilmenau analyzed data continuity for change-process-focused PLM-ERP coupling), roughly 80% of relevant “corporation items” are information items tied to release, configuration, and change management.
But, and this is the critical point, a true digital thread requires more than linked data objects. It requires:
Interdisciplinary structural integration across mechanical, electronics, and software. Linked processes such as change management and quality management connected end-to-end, not just data references. And supply chain integration, with the Asset Administration Shell as the external link to suppliers.
On the architecture side, Martin was blunt: relational databases are unsuited for the digital thread. Graph databases are the right foundation. The digital twin then connects to the physical world via the digital thread — what he called the “Digital Twin Trinity.” And on top of it all sits an AI orchestration layer: select one object, and the system surfaces all connected potential-effect items across disciplines, with AI filtering and acting on them.
If this sounds familiar to readers of Beyond PLM — yes, this is a knowledge graph architecture with semantic context and AI reasoning on top. More on the convergence below.
AI: Evolution, Revolution, and the Agentic Leap
Martin positioned himself carefully on AI. Of the three tiers — narrow AI (ANI), artificial general intelligence (AGI), and artificial superintelligence (ASI) — his focus is squarely on narrow AI for practical engineering and PLM tasks. But he admitted something I found refreshing from someone with 40 years of pattern recognition: AI is moving much faster than he ever expected.
The real leap, in his view, is agentic orchestration — moving from single AI agents to coordinated multi-agent workflows. He specifically cited what modern agentic tools (he mentioned Claude’s multi-agent capabilities) can already do, and described his current research: orchestrating the engineering change management process with AI agents.
His benchmark for any engineering company remains what he called his dream for decades — a continuous, holistic, interdisciplinary change management process across the entire product lifecycle. If you want to evaluate the digital maturity of a company, ask one question: do they have integrated, interdisciplinary change management? Most don’t.
He was also honest about limits and risks. The NVIDIA Omniverse demo of BMW’s factory planning in Hungary is impressive and completely out of reach for SMEs today. And on IP: product knowledge and “product memory” must run on single-tenant, private infrastructure. You cannot send your intellectual property to a shared consumer AI service and hope for the best.
Full presentation is available via this link – Martin Eigner 40 Years of PLM Keynote.
The Vendor Question: Open or Closed?
The Q&A produced the most quotable exchange. Asked what PLM vendors will say about this trajectory, Martin predicted a split: some vendors will close their interfaces and he pointed to SAP putting its API behind a paywall just weeks before the conference, while others will open up. His longer-term prediction was more radical: the future may not need large proprietary data models at all. Open, licensable data models could replace them, with vendors competing on innovation on top rather than on data lock-in.
And in a moment that got a laugh but was entirely serious, he described his plan to ask an AI coding agent: “Can you create a bill of material?” and see how far modern AI gets toward building engineering software from an open understanding of what a BOM is. Whether or not that produces commercial software, the direction of travel is clear.
His closing thesis: the engineering company of the future is a human-AI orchestration — specialized AI agents for change management, quality, and design, standing on the foundation of a semantic digital thread.
What Is My Conclusion?
Here is what struck me most. Martin and I prepared our keynotes completely independently, coming from different histories – his from four decades of PDM/PLM system building and academic research in Germany, mine from CAD data management, BOMs, and building OpenBOM. And yet the destinations converged: graph-based semantic architectures instead of relational monoliths; connected context instead of siloed records; AI agents orchestrating engineering processes instead of humans pushing workflows through 30-year-old systems; and openness of data models as the battleground of the next decade.
As Martin said at the start of his talk – we may have slightly different views, but the final direction is the same. When a field’s pioneers and its newer critics independently arrive at the same architecture, pay attention. In my next article, I’ll share my own keynote — “From CAD Files to Product Memory” and the book announcement that came with it. Just my thoughts…
Check the next article is coming.
Best, Oleg
Disclaimer: I’m the co-founder and CEO of OpenBOM, an AI-native collaborative digital thread platform connecting engineers and manufacturing teams. My opinion can be unintentionally biased.
