I just came from Jerez (Spain) where I attended Share PLM Summit 2026. In my article today, I wan to share my reflections from the event, why AI needs memory context, why Product Memory depends on people, and how companies can start building AI-enabled workflows today.
I joined Martin Eigner and Helena Gutierrez at the AI workshop and one of the opening questions was: “How often do you use AI?” Almost every hand in the room went up.
That moment changed something in me. Not because it was surprising. I actually expected AI adoption to be widespread, but because of what it meant for every conversation that followed. We no longer need to convince people that AI matters. We no longer need to spend time defending the idea that something fundamental is happening. That argument is over. Everyone is already using it.
The more interesting question starts exactly there. But before I move to discuss my reflection, I want to start from the event itself. Share PLM Summit was one of those events where the formal agenda was only part of the experience. The real value came from conversations, questions, discussions, dinners, workshops, panels, and the particular energy that forms when people who care deeply about product development, engineering, manufacturing, and PLM come together in the same place.
It all starts from the place. Michael Finocchiaro wrote “100+ PLM professionals walked in the Bodega in Jerez…” – that starting point already defines everything. The building full of winery fumes creates a different atmosphere for the event.
AI was not just a keynote topic. It was not a separate track. It was not something people mentioned at the end of a conversation. AI was in every conversation. It was in discussions about PLM strategy, data, implementation, business transformation, organizational change, and the future role of people.
But the raised hands were the real signal. The adoption question is behind us. What comes next is harder and more interesting: how do we bring AI from personal productivity into business processes? How do we make it useful for engineering, manufacturing, supply chain, service, and decision-making at scale? And perhaps the most important question of all: what is the role of humans now?
This was my biggest takeaway from the Share PLM Summit. The AI conversation is quickly moving beyond tools and prompts. It is becoming a conversation about product context, organizational memory, human relationships, and business value.
In other words: AI needs Product Memory. And Product Memory needs people.
AI Adoption Is No Longer the Question
For the last two years, many AI conversations started with curiosity. People were asking what tools to use, what prompts work, and how to experiment with generative AI. These were important questions, but they mostly belonged to the first phase of adoption.
At Share PLM Summit, I felt that this phase was already behind us. People are using AI. They use it to write, summarize, translate, search, analyze, code, prepare presentations, and explore ideas. AI has become part of the daily work routine for many professionals.
But personal usage of AI and business transformation with AI are not the same thing.
When I use AI personally, I bring the context. I know what I am working on. I know what document I am editing. I know what documents and other pieces of information I bring in. I know whether the answer makes sense. I can judge the output because the context is in my head.
Business processes are different. The context is not in one person’s head. It is distributed across CAD files, BOMs, ERP systems, spreadsheets, supplier records, emails, quality reports, service history, change orders, and thousands of decisions made over time. Some of this knowledge is structured. Some of it is hidden in documents. Some of it lives only in people’s experience.
This is why the business AI question is much harder than the personal AI question.
The problem is not access to a large language model. The problem is to access and organize a product context. Without context, AI remains a smart assistant looking at fragments. With context, AI has the potential to become part of a new operating model for engineering and manufacturing work.
The Second AI Question: What Is the Role of Humans?
A lot of AI discussions start with technology. What model should we use? What tools should we deploy? What can be automated? These are important questions, but they are not sufficient.
There is another question that is just as important: what is the role of humans now?
This question came up again and again in Jerez. It was not always asked directly, but it was present underneath many conversations. If AI can summarize, analyze, recommend, and generate, what should people do? How should work change? How should organizations change? How do we preserve responsibility, judgment, creativity, and trust?
Manufacturing companies are full of decisions that cannot be reduced to simple automation. Engineers, manufacturing planners, supply chain managers, service teams, and executives are constantly making trade-offs. Cost, quality, availability, compliance, customer commitments, engineering intent, manufacturing capability, and long-term product strategy all collide in daily decisions. No algorithm resolves those trade-offs on its own — not because the algorithm lacks capability, but because the answer always depends on context, consequence, and accountability that lives with people.
AI can help with many tasks. It can find patterns, compare data, identify missing information, summarize changes, detect inconsistencies, and propose options. But humans remain essential because humans understand consequences, relationships, and timing in ways that extend far beyond what any current system can represent.
This is why the human question is not philosophical. It is very practical. Helena Gutierrez shared more about it in her keynote as a three steps – automate the commodity; build the enterprise harness, and elevate the human premium.
The outcome of the process is how AI will industrialize intelligence production.
The more AI enters business processes, the more important human judgment becomes. The role of people shifts — from chasing data and manually reconstructing context, to asking better questions, making better decisions, and validating AI-assisted recommendations. In this sense, AI does not remove the need for people. It changes where human value is created.
Everyone Agrees Data Matters. The Hard Question Is How.
Another theme that was clear at the summit was data. Everyone understands that AI needs data. This part is no longer controversial. The harder question is how to organize data so it can actually support AI-enabled workflows.
Manufacturing companies have been managing product data for decades. They have CAD systems, PDM systems, PLM systems, ERP systems, MES systems, document management systems, spreadsheets, and many other tools. Each system manages a piece of the product lifecycle. Each system has its own logic, its own data model, its own ownership, and its own version of truth.
The traditional PLM answer was built around control. Define the process. Manage revisions. Approve changes. Protect released data. Control access. Maintain traceability.
All of this is still important. Manufacturing companies still need configuration management, compliance, revision control, and traceability. But control is not enough for the AI era.
AI needs more than approved records. AI needs context. It needs to know not only what the product is, but how it got there. Why was this component selected? What alternatives were considered? What supplier problems happened before? What manufacturing feedback came through? What service issues appeared later? What changed between revisions, and why? Who made the decision, and what assumptions were in place at the time?
Most current systems were not designed to capture this kind of context continuously. They were designed to manage official data states and formal processes. The informal knowledge around those processes — the real story of how a product developed — often disappears into meetings, emails, spreadsheets, and people’s memories.
This is the gap that companies are now starting to recognize. The question is no longer simply “Do we have the data?” The question is: do we have the right context for AI and people to make better decisions?
My best example came from the keynote of Antonio Casaschi from ASSA-ABOLY Group – traditional enterprise systems will play a secondary role in future enterprise landscapes.
From Data Control to Decision Impact
One of the most important shifts I see in PLM right now is the move from data control to decision impact.
For many years, PLM was mostly discussed as a system for controlling product data and managing processes. This was logical. Companies needed to prevent mistakes, manage revisions, coordinate engineering changes, and ensure that people were working from correct information.
But the AI conversation changes the center of gravity.
Control asks: is this the approved revision? Decision impact asks: is this the right decision based on everything we know?
Control asks: who approved the change? Decision impact asks: what are the consequences of this change across engineering, manufacturing, supply chain, and service?
Control asks: is the process followed? Decision impact asks: can the organization learn from what happened before and make a better decision next time?
I am not suggesting control disappears. It does not. In many industries, control is mandatory. But control alone does not create full business value. The value comes when product data becomes active, connected, contextual, and genuinely useful for the decisions people make every day.
This is where AI can become powerful. But only if it has access to product context.
The confirmation of this approach came from multiple presentations. Here are my two examples from Jos Voskuil keynote and presentation of Marcellus Menges from Sick AG
Which brings me to the conversation about missing layer to support new organization of workflows and tasks.
Product Memory as the Missing Context Layer Beyond PLM
Product Memory came up in many discussions at Share PLM Summit. I found that people immediately understood the problem behind the concept, even when the terminology itself was still new to them.
Companies know that product knowledge is fragmented. They know that important decisions are lost over time. They know that engineers spend too much time searching for information. They know that manufacturing often needs to reconstruct engineering intent. They know that supplier, quality, and service knowledge is disconnected from design decisions. They know that when experienced people leave, part of the organization’s memory leaves with them.
Product Memory gives a name to this missing layer.
It is not just another database. It is not a document repository. It is not only a PLM workflow. It is not simply a digital thread, although it is related. Product Memory is the organized context of a product: what it is, how it changed, why decisions were made, who was involved, what alternatives existed, and what happened across engineering, manufacturing, supply chain, and service over the product’s lifetime.
Traditional environments we have in almost every organization doesn’t provide a foundation for organization of new types of AI-enabled workflow. The way organizations are working is changing already today (if everyone is already using AI as a personal tool, it immediately impacts the way people work).
Traditional architectures like in the picture below present challenges.
And it reframes old single source of truth concepts PLM was built upon for many years to move it beyond PLM.
This memory connects product structures, files, BOMs, changes, suppliers, decisions, conversations, lifecycle events, and history. It creates a context layer that can be used by both people and AI agents.
This distinction matters. AI does not need more disconnected data. AI needs memory – an organized context that can be used. It needs structured and contextual knowledge that explains relationships, decisions, and changes over time. Without Product Memory, AI is mostly a tool operating on fragments. With Product Memory, AI can become a genuine participant in the product lifecycle — because it has context to reason from.
Human Relationships Are the Original Product Memory
One of the most important realizations I had in Jerez was that Product Memory is not only a technical concept. It is also a human one.
In every manufacturing organization, a huge amount of product knowledge lives in relationships. People know who to call. They remember why a supplier was selected. They know which engineer understands a specific subsystem. They know which manufacturing planner can explain a production workaround. They know what happened in the last project, even if it was never fully documented.
We often call this tribal knowledge, usually in a negative way. And yes, tribal knowledge can be a serious problem when it is hidden, fragile, and dependent on a handful of specific people. But tribal knowledge also has a positive side. It reflects trust, experience, shared history, and practical understanding that formal systems rarely capture.
The goal should not be to eliminate tribal knowledge. The goal should be to transform it into Product Memory.
This is where human relationships and Product Memory intersect. People create context. People explain why things happened. People connect facts with meaning. People understand the difference between a formal approval and a real decision. People know the story behind the data.
Product Memory should not replace this human layer. It should preserve it, connect it, and make it available over time — so that human knowledge becomes more durable, more transferable, and more useful for both people and AI agents.
The Share PLM Summit itself was a strong signal of this. In the middle of all the AI discussions, the event demonstrated the continuing importance of human relationships. People came to Jerez not only to hear presentations. They came to meet, discuss, challenge, listen, and build trust. That did not feel like something AI was making less necessary. It felt like something AI was making more necessary, precisely because the judgment and relationship layer becomes more valuable as AI handles more of the operational layer beneath it.
I want to say a special thank you to the Share PLM team for creating this event and for the enormous amount of work that goes into making it happen. Events like this do not organize themselves, and the quality of the conversations in Jerez reflects the quality of the community they have built over time. What they have created is genuinely rare in this industry.
And this brings me to something I think about often. I have had many conversations about PLM, AI, product data, and the future of manufacturing through podcasts, webinars, LinkedIn posts, and written articles. Those conversations have real value. But meeting people in person is simply different. It is not a matter of degree, but it is a different kind of knowing.
During the past few months I attended episodes of the Future PLM podcast by Michael (Fino) Finocchiaro and those conversations were good. But having the opportunity to get the entire group of webinar participants together in Jerez, having dinner, walking between sessions, and talking without an agenda produces insights and connections that a recorded conversation never could.
Thank you Rob Ferrone for the original idea of the picture (better than placing a gold chain of Per-Chasing on my neck like you did at Share PLM Summit 2025) and amazing group discussing PLM ideas – Martin Eigner, Jos Voskuil, and Patrick Hillberg, PhD.
Something shifts when you are in the same room. People say things they would not write. They respond to the energy in the moment. They build trust in a way that a video call or a comment thread cannot replicate. That is the irreducible value of in-person events, and it is exactly the kind of human context that no system (AI or otherwise) has yet found a way to replace.
Business Value Comes from Remembering Better
The business value of Product Memory is not abstract. It comes from solving a very practical and often invisible problem: companies spend enormous time and energy reconstructing context they have already created, over and over again.
Every time someone asks why a part was selected, why a change was made, what changed between two revisions, what supplier issue happened before, or what manufacturing impact a design decision might create, the organization is trying to rebuild memory it once had but never preserved.
This cost is everywhere.
It appears when engineers repeat old mistakes because they cannot find the previous rationale. It appears when manufacturing teams need to clarify engineering intent before they can begin production. It appears when supply chain teams do not understand which components are critical and which can be substituted. It appears when new employees need months to understand product history that should be accessible in hours. It appears when change review meetings run long because the context must be manually rebuilt from scratch before anyone can evaluate the decision.
One of the biggest hidden costs in manufacturing is the cost of reconstructing context. And because it is hidden — because it appears as meeting time, email threads, and individual effort rather than as a line item — it almost never appears in the business case for technology investment.
Product Memory reduces this cost. It helps companies remember better. And when companies remember better, they decide better.
This is where AI can create real business value: not by answering generic questions, but by helping people make better decisions inside real workflows — reviewing BOMs, detecting missing information, comparing revisions, summarizing change impact, identifying inconsistencies, preparing handoff packages, helping new team members understand product history faster. All of this becomes possible when AI has access to organized product context.
What To Do Next: How To Start Your AI Project
The practical question many companies are asking now is simple: where do we begin?
My recommendation is not to start with AI technology. Do not start with a chatbot, a model evaluation, or a vendor demo. Start with your business reality. AI will only create value when it has enough context to support real work.
The first step is to understand your data. Product knowledge today is scattered across CAD files, BOMs, spreadsheets, ERP systems, supplier records, quality reports, service data, emails, and people’s experience. Before asking AI to help, you need to understand where this information actually lives, how it changes, who owns it, and where the most painful gaps are. Which data is structured? Which is duplicated? Which is trusted? Which is locked inside files? Which knowledge exists only in specific people’s heads?
This does not mean creating a perfect single source of truth before starting. That would be unrealistic, and frankly unnecessary. The goal is to understand the current state and identify where context is missing. That map is your starting point.
The second step is to analyze workflows. Where do people spend time searching, comparing, validating, reconciling, or explaining product information? Where do engineers and manufacturing teams lose time reconstructing context? Where do change processes slow down because the organization cannot quickly understand what changed, why it changed, and what it will impact downstream?
AI opportunities are usually hidden inside repetitive, context-heavy tasks — the places where people spend too much time collecting information, checking consistency, preparing summaries, or explaining product decisions to someone who needs them in a different form.
The third step is to build context. This is where Product Memory becomes important. AI cannot create business value from disconnected fragments. It needs organized context: product structures, relationships, decisions, changes, ownership, history, dependencies. This context does not need to be perfect on day one. Start with one product line, one BOM process, one change workflow, one supplier handoff. The goal is to create enough context for AI to support one meaningful business task.
Only then should you plan AI-enabled workflows. The right question is not “how can we use AI?” The better question is: how can AI improve this specific workflow, in this specific context, for these specific people?
Can AI review BOMs and detect missing information? Can it summarize changes and prepare a review package? Can it compare revisions and explain differences in plain language? Can it help engineers understand supplier constraints before they finalize a design? Can it prepare manufacturing handoff documentation? Can it help new team members understand product history in days rather than months?
These are not generic AI experiments. These are workflow improvements built on product context. And that distinction is everything.
The goal of an AI project is not to demonstrate that AI can answer questions. We already know it can. The goal is to improve business work. That requires starting with data, understanding the workflow, building the context, and only then applying AI where it will actually change an outcome.
Conclusion: The Future Is Human, AI-Assisted, and Memory-Driven
My biggest takeaway from Jerez is this: the future of PLM will not be defined by AI alone, and it will not be defined by better data management alone. It will be defined by the intersection of AI, organized product context, and human relationships — and the companies that figure out how to connect all three will have a substantial advantage over those still treating them as separate problems.
AI is important. It will change how we search, analyze, summarize, validate, and make decisions. But AI cannot create business value in a vacuum. It needs product context. It needs memory. It needs an understanding of relationships, history, decisions, and consequences that has been deliberately organized and preserved.
At the same time, Product Memory is not only a technical architecture. It depends on people. People create meaning. People build trust. People explain why decisions were made. People understand the business reality behind the data. The story behind the approved revision is always a human story.
The room full of raised hands in Jerez told me that the first chapter of AI adoption is already written. Everyone is using AI. That is no longer the interesting question. The interesting question — the one that will define the next several years of PLM and engineering — is whether companies can organize their product knowledge well enough to give AI something real to work with.
The companies that do this will not just use AI better. They will remember better. And companies that remember better make better decisions, move faster, lose less knowledge when people leave, and build products that reflect everything they have learned, not just what they can still find.
That is the real opportunity. And it requires starting not with AI, but with memory.
Oleg Shilovitsky is the co-founder and CEO of OpenBOM and the author of the Beyond PLM blog, where he writes about the future of product development, PLM, and manufacturing technology.
