Shell Deploys C3 AI Agents to Automate Predictive Maintenance

Shell is deploying autonomous AI agents from C3 AI to automate predictive maintenance across more than 30,000 critical equipment assets in its upstream and downstream operations, according to details first reported by AI News.

The expansion builds on Shell's existing use of the C3 AI Reliability Suite, which currently monitors equipment for anomalies. The new agentic AI layer will handle the entire maintenance lifecycle—from initial warning signs through completed repairs—with minimal human oversight.

From detection to autonomous action

Shell's earlier machine learning systems flagged unusual patterns in sensor data and alerted engineers. The new AI agents go several steps further: they independently investigate root causes, draft work orders, verify parts availability in inventory systems, and generate procurement requests.

The agents operate on top of C3 AI's platform, which ingests high-frequency sensor data from operational technology (OT) systems and combines it with business context from ERP platforms like SAP. Each agent is configured for specific equipment—pumps, turbines, compressors—and learns normal operating baselines for that gear.

When the underlying machine learning models detect a deviation, the agent activates and gathers contextual data including maintenance history, environmental conditions, and upstream process variables. It then recommends a fix with supporting evidence. Human operators can approve or override the plan, and as the system proves reliable, Shell can fully automate responses to certain alert types.

"This expanded partnership with Shell proves what's possible when enterprise AI is fully operationalised at global scale for predictive maintenance—reducing unplanned downtime and delivering hundreds of millions of dollars in economic value," said Stephen Ehikian, President of C3 AI.

Why it matters

Many industrial companies can predict equipment failures but struggle with the "last mile"—turning insights into fast action. Engineers typically must manually investigate alerts, determine root causes, and write work orders. By automating these steps, Shell can dramatically shorten the time between predicted failure and actual repair, directly improving uptime and production continuity. The shift to condition-based maintenance also reduces costs by eliminating unnecessary work on healthy equipment and extends asset lifespan. In the energy sector, preventing catastrophic failures before they occur carries significant safety and environmental benefits.

Integration with existing workflows

Direct integration with SAP and other enterprise systems allows the AI agents to operate within the same workflows human planners use. This tight coupling between predictive models and execution systems is what enables true automation rather than just better alerts.

"What Shell and C3 AI have built on Azure over the past several years is exactly what enterprise AI should look like—real applications, running in production, delivering measurable value at global scale," said Sandy Gupta, VP GISV, Software Development Companies at Microsoft.

The deployment represents a shift from algorithmic predictions to practical industrial AI production workflows where the system's value comes from its ability to act on predictions with minimal human intervention.

AI News first reported the details of Shell's expanded C3 AI deployment.