In this recent Beyond PLM article, the author challenges the hype around “AI-native PLM,” arguing that most current offerings amount to AI “wrappers” around old systems rather than genuine next-gen platforms.
Traditional PLM systems were built for deterministic data management: structured bills of materials (BOMs), revisions, workflows, and version control. As it stands, many vendors are simply layering AI features, such as search assistants, chatbots, and auto-classification, on top of those legacy architectures. That may help with user convenience, but it does not make the system truly “AI-native.”
The article draws a line between superficial enhancements and deeper transformation. Real value from AI in PLM shows up only when data is modeled with explicit semantics, workflows are “agentic” (i.e., capable of automated reasoning or decision support), and systems are rearchitected to let probabilistic AI operate alongside traditional deterministic engines. This means a shift toward polyglot data stores, e.g., relational databases for revision control, graph- or vector-based stores for relationships and semantics, and LLM or vector-AI layers for reasoning and search.
Short-term gains may come from targeted AI use cases: automated classification, multi-view BOM analysis, change-order assistance, procurement, and supply-chain risk workflows. These can reduce friction today without forcing a full-scale architectural modernization.
In conclusion, the author suggests treating “AI-native PLM” not as a marketing label but as a direction: one that requires clean data, semantic modeling, connected workflows, and human-in-the-loop governance. True transformation will come not by bolting AI onto existing systems, but by evolving PLM infrastructure so AI becomes a meaningful partner, not just an add-on.