A striking result from recent research at MIT is that about 95 percent of generative AI investments yield no measurable returns. That is a wake-up call. It does not mean AI is useless, but it does mean that most efforts are misdirected. If companies treat AI as a shiny add-on, rather than as a solution to a concrete pain point, they fall into the trap of hype without impact.
In the world of PLM that trap is especially dangerous. PLM systems carry the weight of design data, change processes, compliance, supplier inputs, and versioning. They already live at the intersection of multiple disciplines.
Introducing AI into PLM should not be about branding or calling “We Use AI.” It should be about solving real, painful business challenges like reducing approval cycle time, catching design errors, improving data quality, automating repetitive tasks, or improving decision consistency.
How PLM Vendors Are Adapting AI
Dassault Systemes, with its 3DEXPERIENCE platform, has taken a more ambitious approach by embedding AI into its broader vision of “augmented design” and “virtual twins.” Their AI capabilities support generative design, where the system can propose new design options based on constraints like weight, material, and cost. Dassault has also demonstrated natural language search across its product lifecycle data, helping users navigate complex product structures.
Siemens on the other hand, has been very active in embedding AI into Teamcenter. Their recent releases introduced a Teamcenter Copilot that allows engineers and managers to interact with product data conversationally. Instead of manually navigating complex structures, users can ask natural language questions such as “show me all components in this assembly with pending changes” or “find parts with compliance risks.” This AI assistant uses RAG techniques to ground its responses in the company’s actual product data.
These are real business gains, engineers spend less time hunting for BOM data or instructions, translators and documentation teams spend less time manually rewriting, and manufacturing errors due to misinterpreted instructions get reduced. These AI features are not marketing fluff, they target known pain points.
Aras Innovator has also started weaving AI into its platform, but in a characteristically flexible and open way. For example, Aras has showcased AI solutions where AI agents analyze engineering change orders, suggest risk impacts, and even help generate AML queries dynamically for administrators. Because Aras is built on an open and adaptable data model, AI can be configured to sit directly in the workflows of change management, variant design, or supplier collaboration.
At the Aras ACE 2025 user event, Aras introduced “agentic AI” capabilities more overtly, signaling that its roadmap is focused on embedding more autonomous decision assistants into the PLM layers
Why Real Success is Still Rare?
These vendor examples give promise. But even with them, fear of falling into the 95 percent of AI projects that see no return remains real. The reason is simple, if these capabilities are not anchored to real day to day business problems, they won’t stick.
A copilot that answers casual questions is nice, but if engineers don’t need or trust those answers, they will bypass it. A translation tool is useful, but only if the translations are accurate and reduce rework. An AI suggestion in change management is helpful, but only if it avoids costly errors or accelerates approval cycles measurably.
A Practical Path Forward
The MIT study is a warning, but not a death sentence for AI. For PLM implementations, the message is simple, do not add AI unless it clearly fixes a real problem.
If the engineers spend weeks searching for duplicate parts, then AI should focus on part classification and similarity detection. If the engineering change approval cycles drag on, then AI should analyse change orders, highlight risks, and reduce bottlenecks. If the global teams waste time translating instructions, then AI should automate translations with the right technical context.
AI in PLM will only succeed if it solves real business pain, saves time, reduces errors, and improves collaboration. Anything else is a distraction.
My focus is on helping organizations optimize their product lifecycle processes, enhance collaboration, and achieve sustainable growth through effective PLM strategies. Dedicated to delivering value, I strive to empower clients to overcome challenges and achieve their business goals.