You are exploring how language models can support industrial optimization work, where users need answers and actions grounded in operational knowledge rather than generic model priors. The problem may involve technical documents, operating procedures, asset notes, engineering constraints, and multi-step workflows that connect analysis to recommendations.
How would you apply LLMs, generative AI, and agentic AI workflows to industrial optimization problems?
A representative use case is an engineer asking why a furnace is losing efficiency, with evidence spread across shift notes, maintenance logs, historian summaries, and operating procedures. The system must retrieve the right evidence, synthesize likely causes, cite sources, and avoid unsupported control recommendations.