AVEVA Shapes Industrial AI: Humans in Control

AVEVA formalizes a governance framework that places safety and human oversight at the heart of AI deployments in industry. Arti Garg discusses ongoing methodologies at IEEE to measure the environmental footprint of AI and describes concrete applications ranging from data correlation to robotic intervention. Adoption has progressed significantly, despite models being deemed less predictable.

Measuring AI's Footprint and Maintaining Human Accountability

AVEVA believes that organizations need to better understand the environmental footprint of AI. Arti Garg is involved in an IEEE working group that is developing a standard method to assess the impact of AI on electricity, energy, resources, water, and carbon. This initiative is currently underway. The company emphasizes a threefold mandate for responsible AI: safety, efficiency (including environmental), and prioritizing human safety and oversight. Arti Garg stresses the importance of keeping humans at the center of decision-making, with safeguards defining the areas of action for automated systems and those where supervisors remain accountable. AVEVA's framework thus prioritizes safety, efficiency, and human security, with effective oversight.

Autonomy in the Field: Robots and Diagnostics Without Operator Exposure

The next step for industrial AI could see an increased presence of autonomous robots and drones in physical environments. Arti Garg anticipates that these autonomous systems will transform work in factories, energy systems, and mining sites, making operations more efficient, safer for humans, and more productive. Robots are described as capable of moving, collecting data, and enabling rapid diagnostics, either through embedded computing or in connection with an operator, thus avoiding exposing the latter to hazardous environments. These robots could also gather information without direct intervention from workers.

From Data Chaos to Real-Time Support for Operators

In industry, data comes from various sources such as telemetry, service logs, or engineering documents, and must be correlated. This correlation work is deemed essential for leveraging digital technologies. Arti Garg cites the example of assembling information on a pump or mixer from telemetry, intervention histories, and design and maintenance documents. Graph databases and AI allow for faster association of these disparate sets. The goal is to provide near real-time support to operators during an incident, with the possibility of using a tablet to retrieve, correlate, and diagnose information.

AVEVA Applies a Common Framework to Internal AI and Customer Products

AVEVA describes a multi-level governance structure covering both the internal use of AI and its integration into customer-facing products. A similar framework is applied in both cases to manage deployments. This approach aims to regulate the use of AI while leveraging its benefits. Arti Garg asserts that AI should complement humans, not replace them, in critical decision-making loops.

Accelerating Adoption Despite Less Predictable Models

According to Arti Garg, the last two years have seen significant progress in physical AI, which accelerates robotics and autonomous systems, while agentic AI supports software automation. General-purpose foundational models are broadening access to advanced techniques. Despite past hesitance related to unpredictability of outcomes, adoption is accelerating in industry. The adoption of industrial AI has increased by 78% over the past two years, which Garg considers a turning point. However, she emphasizes that these new types of AI remain less predictable in their behavior.

Goal: Safer, More Efficient, and Sustainable Automation

Industrial AI is presented as entering a new phase, driven by foundational models, physical AI, and agentic AI, but facing direct interaction with physical systems where an unforeseen decision can impact safety, reliability, and critical infrastructure. Responsible deployment is highlighted. AI-assisted programming is cited as a means for domain experts to design new applications. Arti Garg believes that realizing these prospects requires rethinking processes, establishing appropriate safeguards, and providing experienced personnel with new ways to express their expertise, aiming for more autonomous, safer, more efficient, and sustainable automation. AVEVA, which has been working on industrial AI for over 20 years, focuses on critical operation environments where human safety is a key factor and seeks to reconcile technological promise with maintaining safety and reliability.