This session addresses the challenge of enabling LLMs and SLMs to reliably work with industrial engineering and operations data. We’ll cover approaches that avoid full model retraining, effective RAG patterns for accessing high‑value data, and when fine‑tuning is warranted. We’ll also outline how to combine these methods in a resilient production architecture, manage agent tool orchestration, maintain data relevance, optimize grounding, and anticipate failure modes to deliver accurate, context‑driven AI integrated with existing engineering workflows.