The rapid adoption of artificial intelligence is reshaping how engineering organizations manage product knowledge, prompting new questions about the role of “Product Memory.” In this Beyond PLM blog article, Oleg Shilovitsky argues that Product Memory should not be viewed as another system of record. Instead, it represents a new architectural layer that preserves engineering reasoning across existing enterprise systems. While traditional Product Lifecycle Management (PLM), Enterprise Resource Planning (ERP), and Computer-Aided Design (CAD) platforms remain authoritative sources for specific data, they often fail to capture the context and rationale behind engineering decisions.
The article traces the evolution of engineering data management over the past four decades. Product Data Management (PDM) systems solved the problem of managing CAD files, while PLM expanded to control product structures, engineering changes, manufacturing processes, and lifecycle information. Although these systems excel at recording approved decisions and maintaining audit trails, they require information to fit predefined object models and workflows. As a result, informal discussions, rejected alternatives, assumptions, and lessons learned often disappear once a project is completed, even though they may prove critical years later.
Shilovitsky compares this limitation with recent advances in AI. Modern AI agents no longer attempt to keep all information in memory. Instead, they retrieve relevant knowledge when needed through indexed connections. He argues that Product Memory follows the same principle by preserving relationships, history, semantics, and attribution across multiple engineering systems rather than storing another copy of enterprise data. This allows engineers and AI agents to reconstruct the reasoning behind decisions instead of relying solely on isolated facts recorded in different repositories.
According to the article, Product Memory complements rather than replaces existing systems of record. CAD continues to own geometry, PLM manages product structures and change processes, ERP governs business transactions, and Manufacturing Execution Systems oversee production. Product Memory does not compete with these systems for authority. Instead, it links them through a graph of knowledge that captures why decisions were made, who made them, and what evidence supported them. As engineering organizations increasingly deploy AI-powered assistants, preserving this reasoning may become as important as preserving the product data itself, enabling more informed decisions throughout the product lifecycle.