Before software entered the engineering room, something else was already there — a system made of paper, people, and presence. Understanding it may be the most important thing we can do to prepare for what AI brings next.
My first memory of engineering
I probably started learning about engineering before I knew the word.
My father was a mechanical engineer at a metallurgical factory in the Soviet Union. He led a group of designers. As a child I sometimes visited him in the office, and I still remember the atmosphere of that place: large drawing tables, sheets of paper, pencils, rulers, and people standing around the work, discussing, commenting, suggesting corrections.
What stayed with me was not the drawings themselves. It was the way people worked around them.
My father would walk from one drawing table to another. He would ask questions, point to details, challenge assumptions, suggest changes. Sometimes the correction was small. Sometimes it was conceptual. But the process was always human, built on experience, conversation, judgment, and memory.
The drawing was the object on the table, but the real work was happening in the conversation around it.
I also remember his draft sketches. Today people say “back of a napkin” to describe the first rough version of an idea. But in the Soviet Union, napkins were not exactly abundant. Newspapers, however, were everywhere, printed in enormous quantities, full of the same official propaganda. So newspapers became the convenient surface for ideas. A mechanism, a fixture, a quick thought, all could appear on a newspaper margin before they ever became a formal drawing.
Long before an idea becomes a document, it is a fragile piece of human memory looking for a surface.
When I was in school, we had drawing lessons, and I loved them. Making a drawing was almost magical. You could take something that existed only in your imagination and preserve it on paper. A line was not just a line. It was a way to save an idea, make it visible, and allow someone else to understand what you had in mind.
Only much later did I understand what I was really watching in my father’s office. I was watching product memory, before software, before PLM, before any of the systems we use today. The knowledge of a product did not live in one place. It moved between drawings, conversations, comments, corrections, and the experience of people who had been there long enough to know the story behind the document.
When efficiency is not enough: what Starbucks and vinyl records tell us
A good article caught my attention recently, shared by Helena Gutierrez on LinkedIn. The idea was simple and powerful: as AI makes production cheaper and more abundant, value shifts toward the things technology cannot easily replicate. Human connection. Meaning. Care. Identity. Experience.
Maybe the future of work is not less human. Maybe it becomes more human than ever.
We can see signs of this shift already. Starbucks, after years of optimizing operations, mobile orders, and store throughput, paused the rollout of its automation program and invested more than $600 million in putting more people back in its stores. CEO Brian Niccol has been explicit that craft and connection cannot be achieved with optimized equipment the same way they can be achieved with hospitality. He calls AI a co-pilot, not a replacement.
Then there is vinyl. Streaming made music instantly available, searchable, portable, and nearly free at the point of listening. From a pure efficiency standpoint, vinyl makes no sense. You need a physical record, a player, careful handling, and a ritual. But that is exactly the point. According to the RIAA, U.S. vinyl sales passed one billion dollars in 2025, the first time since 1983, growing for nineteen consecutive years.
Human experience is not only about efficiency.
When I think about those two examples together, I keep returning to my father’s drawing office. Because what I witnessed there was not inefficiency. It was presence. It was shared attention. It was the kind of coordination that only happens when people are gathered around the same object, looking at the same thing, bringing different kinds of knowledge to bear at the same time.
Before CAD: engineering knowledge lived in people, not systems
Before CAD, engineering organizations still had to answer the same questions we ask today.
Which version is correct? Who approved the change? Why was this material selected? Which drawing went to manufacturing? What did the shop floor discover that engineering never heard about?
The tools were different, but the coordination problem was identical.
Drawings were created by hand. Revisions were marked manually. Changes moved as paper forms. A senior engineer remembered why one material was selected over another five years ago. A manufacturing planner knew which supplier could actually deliver on time. A buyer knew which component always caused trouble. A technician knew that the drawing was correct on paper but incomplete in practice.
This was product memory. It just had no name.
It lived in drawings, folders, notebooks, filing cabinets, conversations, meetings, and habits. It lived in the experience of people who had been there long enough to know why a decision had been made. The system worked not because documents were perfect, but because people completed the system. The drawing did not contain all the knowledge. The binder did not contain all the context. Humans carried the missing information across the gaps.
The drawing review was a ritual. It was inefficient by modern standards. People had to be in the same room. Copies had to be distributed. Redlines had to be collected. Someone had to update the master. But the process created something valuable: shared attention, distributed context, and accountability that lived in people, not just in documents.
The drawing was never the whole product story. It was the place where the product story was discussed, corrected, and remembered.
What CAD digitized, and what it left behind
CAD changed the artifact. It replaced manual drafting with digital geometry. PDM then managed files, versions, and releases. PLM expanded further into lifecycle processes and enterprise coordination. Each step digitized another layer of engineering work.
But the human coordination problem remained.
The same questions persisted. Why was this change made? What context was lost between engineering and manufacturing? What decision was made in the meeting but never entered into the system? Which supplier constraint influenced the design?
Digital systems made information easier to store. They did not automatically make it easier to understand. The reasoning behind decisions often stayed outside the formal record, in emails, chat messages, meeting notes, and the memory of people who had been in the room.
Every engineering organization has a hidden layer of memory. It is the senior engineer who knows which part should not be changed. The manufacturing person who remembers what went wrong last time a tolerance was tightened. The buyer who knows which supplier is reliable despite what the approved vendor list says. When these people leave, context disappears. The organization still has data. But it loses memory.
Why this history matters now: AI needs context, not just data
AI makes this problem urgent.
AI agents can read files, summarize documents, compare bills of material, recommend changes, and automate many routine tasks. This is powerful and real. But AI agents need context to be useful. Experienced engineers can work with incomplete information because they bring memory and judgment. They know what the system does not say. AI does not have that background unless we give it to them.
If product data is fragmented, disconnected, or missing decision history, AI will have data without understanding. It will find the file but miss the meaning. It will summarize the document but not grasp the trade-off behind it.
AI does not make Product Memory optional. It makes it urgent.
And this brings us back to those drawing tables.
Before software, humans carried the missing context. People completed the system. They remembered what documents did not say. They connected drawings to manufacturing reality, supplier constraints, customer needs, and past experience. The goal of the next generation of engineering systems is not to remove this human layer. It is to preserve it, augment it, and make it available to both people and machines.
The book I am writing, Beyond PLM: From CAD Files to Product Memory, begins with these ideas. Not with AI. Not even with CAD. It begins before CAD, in the world of drawing tables, paper forms, filing cabinets, and person-to-person coordination. Because to understand where engineering is going, we need to understand what it was before software entered the room.
Before CAD, Product Memory already existed. It was made of paper, people, and presence.
Now we have a chance to make it visible, connected, and useful for the future.
Just my thoughts.
Best, Oleg
Disclaimer: I’m the co-founder and CEO of OpenBOM, a collaborative digital thread platform that helps engineering and manufacturing teams work with connected, structured product data, increasingly augmented by AI-powered automation and insights.
