ABB laid out a six-stage roadmap for industrial autonomous operations at the ARC Industry Forum in Bengaluru on July 9-10, 2026. The framework moves from connecting and contextualizing data to predictive, assistive, agentic, and eventually self-optimizing systems, with Level 4 autonomy-where machines operate under human supervision-identified as the practical target for most industrial applications.
The gap between automation and autonomy
Traditional industrial automation works well when conditions are known and responses can be pre-programmed. Autonomous systems must perceive their environment, interpret changing conditions, and respond to situations that were not explicitly coded. Rajesh Ramachandran, Global Chief Digital Officer for Automation at ABB, told attendees that organizations cannot jump directly from automation to full autonomy.
"At Level 4, the system is mostly in control, while humans continue to supervise operations and intervene when required," Ramachandran said. Full autonomy, where humans are absent from the operational process, would require legal, ethical, safety, and governance frameworks that do not yet exist at scale.
From copilots to agents that act
Industrial software has long served as a system of insight-showing operators equipment conditions and production performance. Generative AI added an assistive layer, with copilots that summarize equipment data, support root-cause analysis, and transfer information between shifts. Those copilots are now evolving into AI agents capable of interacting directly with machines, software, and processes to execute defined workflows.
ABB's recommended progression starts with requiring human approval before any agent takes action. Over time, lower-risk workflows can shift toward greater automation. The level of autonomy granted depends on operational risk, process integrity, and safety requirements.
Six stages toward autonomous operations
Ramachandran outlined a practical sequence that industrial organizations can follow:
- Connect: Gather industrial data from assets and operations.
- Contextualize: Add industrial knowledge and operational context.
- Predict: Use analytical AI to anticipate issues and performance changes.
- Assist: Use generative AI to provide relevant insights.
- Act: Use agentic AI to execute defined workflows.
- Autonomous: Enable self-optimizing operations.
Collecting data is no longer the primary bottleneck. The harder problem is making that data interpretable within its operational context-against an asset hierarchy, production process, safety requirement, or energy target. Without that contextual intelligence, AI systems cannot safely recommend or execute actions.
Industrial AI already in the field
ABB showed several working examples. In asset performance management, AI copilots give maintenance engineers and reliability teams relevant equipment information without forcing them to navigate multiple applications. Digital twins combined with conversational AI help service engineers investigate potential gas analyzer leaks and access repair instructions from ABB's service knowledge base.
On the shop floor, technicians can scan a QR code on a device to pull real-time data and connect with an AI-based service assistant. ABB said its generative AI-powered device management approach resolves up to 80 percent of technical support issues, with remote visual assistance covering many of the remaining cases. These capabilities reduce equipment downtime, cut unnecessary travel by specialists, and support safer operations.
You can watch Ramachandran's full presentation, "The Next Industrial Frontier: AI-Driven Autonomous Operations," on YouTube.
Why this matters for operations professionals
The shift described by ABB changes the operator's job rather than eliminating it. AI handles repetitive analysis, retrieves information, coordinates workflows, and executes approved lower-risk actions. Operators and engineers focus on supervision, exception management, and decisions requiring experience and accountability. For operations managers building these capabilities, the practical takeaway is clear: establish a contextualized data foundation first, then define precisely where AI can assist, act, or optimize. Organizations that skip the data context step will struggle to move beyond basic automation. Those looking to build these skills systematically can explore an AI Learning Path for Operations Managers or review resources on AI for Operations.
Your membership also unlocks:
