Building a safer path to autonomous industrial AI

Industrial AI is on the cusp of a transformative new era, moving beyond traditional predictive analytics and specialized applications. Driven by breakthroughs in foundation models, physical AI, and agentic AI, the technology is now capable of automating increasingly complex tasks across diverse industrial environments. This evolution promises significant gains in efficiency and productivity. However, unlike AI systems that operate solely within the digital realm, industrial AI interacts directly with physical systems—machinery, infrastructure, and operational processes. This direct interaction introduces a critical dimension: an unexpected decision or error by an autonomous system can have severe consequences for safety, operational reliability, and even critical infrastructure. Consequently, responsible deployment is paramount for the success of this next wave of industrial automation. "How do we leverage these technologies while maintaining safety, while maintaining reliable operations, while still being able to deliver on the promises of the new capabilities?" asks Arti Garg, chief technologist at AVEVA. This challenge is particularly acute given that newer AI systems, while more capable, often become harder to predict and explain, complicating oversight. A fundamental pillar for navigating this transition is data. Industrial systems typically generate vast amounts of information from disparate sources, including telemetry, service logs, and engineering documents. Advanced technologies can now rapidly connect and correlate this data, providing operators with real-time insights crucial for diagnosing problems and making informed decisions. Furthermore, AI-powered robots could revolutionize safety by gathering information in hazardous environments, thereby eliminating the need for human workers to enter dangerous zones. Yet, increased autonomy also necessitates robust new approaches to governance. AVEVA's framework for responsible AI, for instance, emphasizes security, operational efficiency, and, critically, human safety and oversight. Garg advocates for AI to augment human capabilities rather than replace people in critical decision loops. This involves establishing clear "guardrails" that define the boundaries of automated actions and delineate where human supervisors retain ultimate responsibility. Sustainability forms another vital component of this equation. AI can play a pivotal role in managing complex power systems, especially as the integration of renewable energy sources expands. Simultaneously, organizations must develop better methodologies to understand and mitigate AI's own environmental footprint, ensuring that technological progress aligns with ecological responsibility. Garg's involvement in an IEEE working group underscores the industry's commitment to developing standards for these critical considerations.