Why do ChatGPT, Claude or Mistral struggle with engineering questions? Because industrial AI has a context problem. In this session recorded at the Tech for Industry Show, Anthony Ponceot (Aras, low-code PLM vendor for 25 years) demonstrates why digital continuity — the digital thread — is the foundation of trustworthy AI on technical data.
The talk covers the fragmentation of product lifecycle data — from requirements (ALM) to design (CAD), simulation, manufacturing and service (CRM) — and Aras' thesis: value doesn't come from the data, but from the connections between them. Without imposing a single tool: the PLM platform links data where it lives. Three concrete cases illustrate the point: a late requirement change (instant impact analysis from requirement to work instruction), component obsolescence (affected variants, alternative parts, compliance impacts — months of resolution saved) and a field issue reported through a serial number (exact configuration, known failures, existing fixes). On the AI side: "thread RAG" — exploring the digital thread to augment the agent's context with the right version, the right variant and the semantics of relationships —, strict enforcement of the permission model by AI agents, and a warning about training models on proprietary data. Closing with 4 implementation requirements and 5 actions to get started.
A reference session for engineering, PLM, quality and IT leaders.
