Arti: For us, responsible AI is guided by a triple mandate: security, efficiency—including environmental efficiency—and, above all, preserving human safety and oversight. These pillars form the foundation of our internal multi-layered governance approach, which applies equally to how we use AI internally and how we deploy it in our products. By leveraging a unified governance model, we ensure that both our own operations and the tools we provide to customers are developed responsibly and effectively.
At the heart of these principles lies a steadfast commitment to keeping humans at the center of AI integration. We view human judgment, responsibility, and ethics as paramount, focusing on AI as a tool to augment rather than replace human decision-makers in critical operational loops.
Megan: That is indeed a critical distinction. As we look toward the next phase of development, many anticipate that agentic AI will enable industrial systems to operate with far greater autonomy. How do the risks associated with AI in physical, industrial environments differ from those in purely digital contexts? Conversely, what benefits and opportunities does software-defined automation present?
Arti: Let’s address the risks first, followed by the opportunities. In industrial environments, the primary risk lies in our interaction with physical systems—systems that are vital to keeping the world running, such as power grids or resource extraction operations. These systems often operate in hazardous environments where equipment directly impacts human safety. When software interfaces with physical systems that have real-world consequences, extreme caution is paramount. At AVEVA, this mindfulness of the end-user and the physical application has been in our DNA from the start, transcending mere on-screen interfaces.
While we have historically utilized AI models by carefully selecting and understanding their behavior—such as our proprietary anomaly detection systems—newer AI capabilities present new challenges. These advanced models are often less explainable than traditional statistical models, and their behavior can evolve over time as they learn. Automating physical systems with capabilities that change dynamically requires profound caution and careful consideration of where and how to apply this increased automation.
However, alongside these risks lie significant opportunities. Just as human experts accumulate knowledge over their careers, these advanced, agentic AI models can learn and gather experience at an accelerated pace. A key advantage is their ability to transfer learned experiences from one site to another, scaling the “learn by doing” principle that defines human expertise across entire operations. This capability to generalize and apply knowledge rapidly represents a profound opportunity for the industrial sector.
The industrial sector is also facing a major workforce transition, with nearly half of the current workforce expected to retire within the next five years. This represents a substantial loss of institutional knowledge and expertise. Capturing this knowledge in actionable formats—and making it accessible to a newer generation of workers accustomed to different technological paradigms—presents a massive opportunity for industrial AI to bridge the experience gap.
Megan: Absolutely. Against the backdrop of these significant market and workforce shifts, the opportunities are immense. Building on this, how can responsibly deployed autonomous industrial AI systems accelerate the transition to faster, more sustainable industrial processes?