Pairing industrial AI autonomy and human control

Digital twin technology in industrial settings has historically trailed physical operations, serving mainly as a mirror for simulation, observation, and what-if planning. Earlier implementations could not perform real-time analysis or act on issues they detected. A newer generation, described as "living digital twins," changes that by combining a dynamic, high-fidelity model of complex operations with the ability to intervene in the physical world through operations and control systems.

AI is central to the shift. Rob McGreevy, chief product officer at AVEVA, compares these AI-infused twins to organisms that grow and evolve: adaptive, evergreen models of a plant's assets and processes that keep pace with live conditions and connect previously siloed systems. That connectivity supplies a layer of visibility and orchestration that earlier setups lacked, better reflecting the rhythms of the facility itself.

Agentic AI pushes the technology further, enabling industrial systems to move beyond diagnostics and suggestions to acting on their own accord. As autonomy advances, the article argues, plants develop something akin to industrial homeostasis, where small, continuous adjustments maintain health and throughput.

Living digital twins connect engineering and design data with real-time operational data through a digital thread — a continuous record of an asset across its lifecycle. Two shifts helped make this possible: cheaper sensors streaming telemetry from IoT devices and industrial control equipment, and software platforms that surface that data in usable interfaces. Predictive and physics-based models add AI-enriched insight into how a machine or process is likely to behave.

Human judgment remains central to the loop. Operators define what "healthy" looks like for a plant and flag when AI systems drift out of alignment; the living twin gives engineers visibility into what the AI is doing and whether execution stays on track. McGreevy frames the twin as a multi-lens tool: an operator can query it for emissions through a sustainability lens, for asset condition through a maintenance lens, or for production volume through a performance lens.

Demand is materializing. A 2026 Manufacturing Leadership Council survey found 66% of manufacturers are using or plan to use agentic AI in operations, and 40% expect some degree of factory autonomy by 2030. Yet risk appetite is tempered: Deloitte's 2025 smart manufacturing survey found 65% of executives ranked operational risk, including business disruption, among their top concerns. What remains to be seen is how quickly manufacturers will hand closed-loop control to AI on the factory floor, which specific processes will be delegated first, and whether the oversight layer humans provide can scale as autonomy widens.