AI-driven automation for industrial digitalization: a scalable framework for network discovery and digital twin deployment

The growing complexity of Industrial Control Systems (ICS) and Operational Technology (OT) networks presents significant challenges in network discovery, device classification, and causal process inference. Traditional methodologies, which depend on manual configurations and static rule-based approaches, often prove inadequate in dynamic industrial environments due to their limited scalability and adaptability. This paper introduces an AI-driven agentic framework designed to automate these critical processes. The proposed system employs autonomous AI agents for real-time network scanning, device identification through communication pattern analysis, and inference of process dependencies. By integrating active and passive data collection into the agents' workflow, where they receive insights from these analyses as input, our approach extracts system dynamics without requiring prior domain knowledge of industrial processes. This methodology advances industrial automation by enabling adaptive, self-optimizing operations, thereby reducing manual intervention and enhancing system visibility. Moreover, it represents a significant step toward the realization of Digital Twins, while also facilitating predictive maintenance and cybersecurity monitoring. Ultimately, this framework offers a scalable and intelligent solution to support the digital transformation of industrial ecosystems.