Artificial Intelligence Tools in Industrial Control: Large Language Models, Agentic Systems and Their Current Deployment

Ing. Dagmar Špičková, Ph.D.¹; doc. Ing. Ivo Špička, Ph.D.²

¹VŠB-TUO, FMT, 17. listopadu 2172/15, 708 00 Ostrava – Poruba, Czech Republic, dagmar.spickova@vsb.cz
²BINTELL SOLUTIONS, Na Vyhlídce 465, 742 85 Vřesina, Czech Republic, ivo.spicka@gmail.com

This paper follows on from last year's review of artificial intelligence tools in industrial control and updates it with key trends from 2025. A fundamental shift has occurred in the area of large language models (LLMs), which have become a standard part of the enterprise environment – notably models GPT-4o, Claude 3.7, Gemini 2.0 and DeepSeek R1. Concurrently, there has been a massive expansion of agentic AI systems, which – unlike conversational chatbots – autonomously plan, reason and execute multi-level workflows. The paper describes key differences between cloud-based and on-premise LLM deployment, approaches to building custom enterprise language systems (fine-tuning, RAG, private models), and methods for automated processing of corporate technical documentation. Specific industrial deployments at companies such as Siemens, ABB and SMS Group are discussed. The paper concludes with perspectives for heavy industry and steelmaking in the context of the EU AI Act regulatory framework.

Key words: large language models, agentic AI, RAG, on-premise LLM, technical documentation

Stáhnout článek