Before starting this article, I want to thank all my colleagues and friends for commenting, sharing their thoughts, and following my articles. Your engagement and participation have been incredibly helpful, and I’m looking forward to building on this momentum and bringing more in 2026.
Now, let’s get back to the topic of this blog.
Sometime in 2025, I caught myself doing something that, honestly, would have felt a bit crazy just a year ago.
I generated a full “PLM strategy” slide deck in a few minutes. Not notes. Not bullet points. A real deck — trends, architectures, maturity models, confident conclusions, the whole thing. Then I generated another one. And another. Ten versions, easily. All of them reasonable. All of them sounding smart. All of them explaining very clearly where PLM is going and what companies should do next.
And that’s when I stopped.
Not because the decks were bad. They were actually good. Too good. The problem was that they were all interchangeable. If you showed me one or another, I honestly couldn’t say which one mattered more. And more importantly, none of them could answer the only question that actually matters in real life: which one is right for my company?
That was the moment something clicked for me.
We live in a time of cheap intelligence. Judgment is quickly becoming a bottleneck…
Explanations are everywhere now. Frameworks are easy. “Best practices” come on demand. Custom GPT tools are impressive — I use them, some of my colleagues built very good ones, and yes, they are useful. They can summarize what already happened, compare options, and generate recommendations faster than any human ever could.
But they cannot decide.
They cannot take responsibility for judgment. They cannot tell you which tradeoff is acceptable for you, in your messy organization, with your politics, your history, your constraints. They cannot tell you which decision you’ll regret less two years from now.
And that realization didn’t make me negative about AI. Not at all. It made me rethink something else.
It made me rethink why I write.
Because if intelligence is cheap, then explanation is cheap too. And if explanation is cheap, then maybe the value is not in explaining anymore. Maybe the value is in judgment.
That realization quietly reshaped everything I wrote on Beyond PLM in 2025.
How 2025 Forced Me to Rethink My Writing About PLM: A Year of Unplanned Conclusions
Going into 2025, I expected to keep writing about PLM helping people to learn more about PLM, technologies, experience, architectures, integrations, and how to use AI to change the way we work – sounded all like the usual topics. The speed of AI development was amazing and I’ve seen how AI keeps changing everything we do.
I sat down earlier today to make analysis of my best articles, trends and reactions. What surprised me was not the topics, but the reaction.
The posts that resonated most were not the ones where I tried to explain how something works. They were the ones where I said, more or less, “Something here doesn’t feel right.” Where I questioned things the industry treats as settled, almost sacred.
That shift — from explaining to judging — was not intentional. It happened while writing, while talking to people, while noticing which posts made people stop and say, “Yes. Exactly. This is what I see too.”
1. From Structure to Meaning
For years, PLM conversations have been about data organization and ownership. EBOM vs MBOM. Single source of truth. Transformations, mappings, synchronizations. We argue about hierarchies as if, once we get the structure right, everything else will magically fall into place, but often overlook the different concepts underlying BOM management. These concepts—such as the distinctions between EBOM, MBOM, and SBOM, and the shift from static to dynamic, model-based approaches—form the foundational framework for managing product data and supporting collaboration across functions.
In 2025, I finally admitted to myself that structure is not the real problem.
The real problem is that we don’t agree on what the data actually means.
You can have two BOMs that are both perfectly structured and still represent completely different intentions. One is about design responsibility. Another is about manufacturing reality. Another is about procurement constraints. We keep trying to align them structurally while ignoring that they encode different semantics.
Once I started looking at PLM problems this way, many old debates suddenly looked strange. Less technical. More human.
And this also explained something that has confused the industry forever: Excel.
People don’t export BOMs to Excel because Excel is better software. They export because Excel lets them think. It relaxes the rules. It allows comparison, comments, negotiation, back-and-forth. Excel is not a system of record. It’s a system of judgment.
Using bullet points or similar formatting, much like Excel’s organizational features, can help clarify and communicate complex PLM ideas and decision-making processes more effectively.
Once I saw that, “Excel leaks” stopped looking like user failure and started looking like system failure.
2. PLM Fails Organizationally
Another uncomfortable thing I realized in 2025 is that PLM rarely fails because it lacks features.
It fails because it avoids ownership.
PLM talks about lifecycle, but it’s often unclear who actually owns decisions at each stage. It talks about collaboration, but quietly assumes everything happens inside one company, even though in reality, collaboration often spans different companies that need seamless integration and communication. It promises executive visibility, but struggles to explain itself in financial or strategic language that a CEO or CFO actually cares about.
ERP didn’t win because it was elegant. It won because it attached itself to money, accountability, and responsibility at the executive level. PLM stayed in engineering, speaking a language that made sense inside engineering, and somehow expected executives to care.
When I started framing PLM problems this way — as questions of ownership and power, not just technology — many adoption stories suddenly made sense. Not acceptable, but understandable. It became clear that we need to openly discuss these issues among all stakeholders to drive real progress.
This also changed how I think about “shared PLM.” Cross-company work is already normal. Ownership is already distributed. PLM systems are just late to admit it.
3. From Systems of Records to Product Memory
One of the strongest threads in my 2025 writing was this idea that PLM is not really a system of record. It is — or should be — a system of memory.
Engineering work is not only about final answers. It’s about why decisions were made, what alternatives were rejected, and what constraints mattered at that moment in time. Most PLM systems are very good at storing outcomes and very bad at storing reasoning.
This is where my thinking about graphs and semantics came together.
A graph is not interesting because it’s modern. It’s interesting because it can represent relationships, context, and evolution. In other words, memory that we need to maintain throughout the entire product lifecycle.
Once you look at PLM this way, a lot of existing behavior suddenly makes sense. Why discussions live in email. Why decisions live in spreadsheets. Why screenshots get passed around in chat tools. People are trying to preserve context, not just data. Capturing what happens in real life business scenarios is crucial, as it bridges the gap between theoretical processes and the actual decisions made day-to-day.
- AI didn’t simplify anything — it exposed the gaps
4. AI Didn’t Simplify Anything – It Exposes Data Gaps
Like everyone else, I spent a lot of time in 2025 thinking about AI. But the more I wrote, the less interested I became in AI as a feature and the more interested I became in AI as a mirror.
AI doesn’t fix fragmented data. It exposes data gaps, making it even more important to manage data and processes effectively to avoid these gaps.
If your data has weak semantics, AI can’t reason over it. If your decisions are scattered across tools, AI can’t reconstruct intent. If your BOMs encode structure but not meaning, AI will hallucinate coherence.
Explanation is no longer scarce. The rise of cheap intelligence means AI and data-driven explanations are readily available and inexpensive, but this abundance makes human judgment and responsibility even more valuable.
That led me to a very simple conclusion: before we talk about intelligent agents, we need intelligent memory.
AI raises the bar. It doesn’t lower it.
- Humor turned out to be a serious tool
Supply Chain Resilience in a Year of Uncertainty
2025 has been a year that forced all my colleagues—and, honestly, most industrial companies—to rethink what resilience really means in the supply chain. Disruptions, shifting markets, and unpredictable events have made it clear that the old ways of managing product data and operations just aren’t enough. The only question is: how do we build a supply chain that can adapt, recover, and even thrive when things don’t go as planned?
This is where product lifecycle management (PLM) has stepped into the spotlight. More companies are realizing that a strong PLM strategy isn’t just about engineering or compliance—it’s about creating a foundation for supply chain resilience. By connecting enterprise systems, leveraging modern PLM tools, and focusing on data management, companies can finally get more control over their supply chain processes.
But let’s be real: PLM implementation is rarely simple, especially for organizations with years of legacy data and deeply embedded manufacturing BOM processes. The challenge isn’t just technical; it’s about changing how teams collaborate, how information flows between suppliers and customers, and how decisions are made in real time.
Digital transformation is the only way forward. Companies that embrace cloud-based PLM solutions and AI tools are finding it incredibly helpful to analyze massive amounts of product data, spot risks before they become problems, and automate routine tasks that used to eat up valuable time. Artificial intelligence and intelligent agents are no longer just buzzwords—they’re becoming essential tools for managing complexity, reducing errors, and improving communication across the entire supply chain.
Product memory is another concept gaining traction. By capturing not just the “what” but the “why” behind every decision throughout the lifecycle, companies can build a single source of truth that supports better planning, faster response to market changes, and more efficient collaboration with suppliers and customers.
5. Humor is a Serious Tool
I was amazed to see how well humor worked… A funny guide for PLM jargon was a great experiment.
When I wrote about PLM jargon in a funny way, people didn’t respond because it was funny. They responded because it was accurate. Humor allowed to people to recognize uncomfortable things without sounding angry or cynical. It lowered defenses.
It taught me something important: people are not tired of thinking. They’re tired of being lectured.
Humor is a way to say, “I see what you see,” before saying anything else.
Blogging itself changed meaning for me
This was probably the most personal realization of the year.
For a long time, writing Beyond PLM meant explaining. Teaching. Laying out frameworks. In 2025, I realized that explanation is no longer scarce. Anyone can generate it now.
Judgment is Scarce
The posts that resonated were not the ones with clean conclusions. They were the ones that named tension and left it open. That admitted uncertainty. That didn’t pretend everything fits nicely.
Beyond PLM slowly became less about being right and more about making sense of what’s happening — together with people who live inside these systems every day. Sharing insights gained from these experiences has proven valuable for informing strategy and improving decision-making throughout the ongoing PLM journey, which is a continuous process of learning and adaptation.
What is my conclusion?
Looking back, here is what I’m taking into 2026…
The biggest thing I learned in 2025 is this – the future of PLM will not be decided by better features or prettier dashboards. It will be decided by whether we can capture meaning, responsibility, and memory in a way that matches how work actually happens. Just as importantly, PLM must be aligned with real business needs and realities to deliver true impact. That leaves me with more questions than answers — and honestly, that feels right.
Where does product knowledge really live today? Who owns decisions when work spans multiple companies? What does it mean to design systems for judgment, not just compliance? And how do we build tools that help captured why we made decisions, not just records?
Implementing these changes comes with significant challenges, from technical hurdles to organizational resistance, especially as companies strive to keep all stakeholders on course during PLM implementation and change management. The cost of poor alignment or missed opportunities can be substantial, affecting both profitability and competitive advantage.
Those are the questions I’ll keep writing about. The introduction phase of new products is critical for success, and manufacturers play a key role in ensuring supply chain efficiency and operational control within the PLM ecosystem. BOM-related services, such as sourcing and configuration, are essential in supporting product development and system integration. Effective PLM also depends on meeting the needs of different users, from engineers to supply chain partners, to foster collaboration and adoption.
Not because I have them figured out, but because pretending they don’t exist no longer feels honest. Ultimately, the value of PLM lies in making informed judgments and decisions that are tailored to each company’s unique situation, not just in generating explanations or frameworks.
Just my thoughts…
Best, Oleg
Disclaimer: I’m the co-founder and CEO of OpenBOM, an AI-native Collaborative Digital Thread platform providing connecting engineers and manufacturing teams.
