Engineering organizations generate and retain enormous amounts of data, yet they can still lose the context needed to understand products and decisions. Beyond PLM’s Oleg Shilovitsky argues that the problem is not lost data but lost organizational memory.
CAD, PDM, PLM, ERP, and manufacturing systems record individual states and transactions. However, important engineering questions often require connecting information across these systems. Understanding why a part is late, for example, may involve its design release, supplier selection, purchase order, lead time, deviation status, inventory, and production schedule.
Shilovitsky calls the missing contextual layer Product Memory. Unlike product data, which describes current or historical states, Product Memory preserves continuity between them. It records what was originally planned, what changed, why decisions were made, and which dependencies were affected. He identifies three important forms: decision memory, operational memory, and dependency memory.
The problem becomes especially visible when digital designs move into physical production. Disconnected information that remained manageable during development can suddenly cause shortages, delays, and line stoppages. Companies often compensate with spreadsheets and experienced employees who manually connect information scattered across systems. As a result, people effectively become an organization’s memory layer.
Better documentation alone cannot solve the problem because engineering knowledge is relational, and some expert knowledge is difficult to formalize. Digital threads connect records across systems, but they do not necessarily preserve the evolving meaning behind those connections.
Artificial intelligence could change this situation. AI systems can access information across engineering and business platforms while capturing changes, conversations, decisions, and relationships as work occurs. Instead of requiring engineers to manually document every decision, AI could reduce the cost of preserving context.
Product Memory, therefore, is not another database. It is a layer connecting product information, decisions, events, people, dependencies, and time so organizations can understand not only what happened but why.
