Pairing Industrial AI Autonomy and Human Control
Living digital twins enable industrial plants to evolve through real-time AI-driven operational synchronization.
1Evolution of Digital Twins
Early digital twins functioned primarily as static simulation tools for observation and scenario planning. Modern living digital twins act as dynamic, high-fidelity models that mirror physical reality in real-time. By integrating operational data with engineering designs, these systems now possess the capacity to actively intervene in complex industrial processes and facility operations.
2Agentic AI Integration
Agentic AI allows industrial systems to move beyond simple diagnostics toward autonomous decision-making. These models develop reflexes for industrial homeostasis, performing constant, small adjustments to maintain optimal plant health. According to a 2026 Manufacturing Leadership Council survey, 66% of manufacturers are currently using or planning to implement agentic AI within their operations.
3Human Oversight Requirements
Despite increasing autonomy, human judgment remains essential for defining healthy operating parameters and monitoring AI alignment. Rob McGreevy of AVEVA suggests that AI will primarily function in an advisory capacity, similar to traditional SCADA systems. Full autonomy is currently limited to narrow, highly repetitive tasks that require minimal complex human intervention.
4Technological Infrastructure Shifts
The convergence of affordable IoT sensors and advanced software platforms has enabled a continuous digital thread across asset lifecycles. Predictive and physics-based models provide AI-enriched insights, allowing operators to view facilities through specific lenses like sustainability or maintenance. This connectivity bridges previously siloed systems, creating a more holistic view of facility performance.
5Operational Risk Concerns
Manufacturers remain cautious about the risks associated with autonomous systems. A 2025 Deloitte survey indicates that 65% of executives rank operational disruption as a top concern. Consequently, organizations are establishing governance boards and principles to manage AI deployment, though many have yet to implement comprehensive safeguards for cross-partner decision-making and accountability.
6Regulatory and Governance Landscape
Regulators are beginning to formalize oversight, with the EU AI Act classifying specific infrastructure AI components as high-risk. Companies are responding by developing internal governance frameworks to audit AI actions. Currently, only 23% of organizations report fundamental changes in accountability, highlighting a significant gap in industry-wide adoption of robust AI governance structures.
7Robotics and Physical AI
The integration of physical AI, such as drones for pipeline inspections, reduces human exposure to hazardous environments. By pairing robotics with AI-powered decision-making, companies can identify gas leaks or hazards more efficiently. This convergence of software and hardware is expected to decrease operational incidents and outages while improving overall facility sustainability metrics.
8Integrated Value Chains
Living digital twins enable the creation of integrated value chains that span entire industrial networks. This connectivity allows companies to react rapidly to supply chain shifts, material availability, and geopolitical turbulence. Achieving this level of resilience is becoming an absolute necessity for maintaining effective business operations in an increasingly complex global market.
Original article: MIT Technology Review — Pairing Industrial AI Autonomy and Human Control