Shell deploys C3 AI agents for predictive maintenance

Shell will deploy autonomous agents from C3 AI on its C3 AI Reliability Suite, which monitors more than 30,000 assets across upstream and downstream operations, to automate predictive maintenance workflows from anomaly detection through work orders, parts checks and repairs.

The existing Reliability Suite ingests high-frequency operational technology sensor data and combines it with business context from systems such as SAP and maintenance logs. Until now, Shell used machine learning mainly to flag unusual sensor patterns for engineers to review. The added agent layer is built to investigate why an alert fired and assemble the contextual evidence needed to recommend or carry out repairs.

Operators will configure agents for individual pieces of equipment by setting objectives and defining permitted responses. When core models detect a deviation from an established baseline for pumps, turbines or compressors, an agent will gather recent maintenance history, environmental conditions and upstream process variables to identify probable causes.

The agent can draft a precise work order, check inventory and generate procurement requests inside the same planning workflows used by human teams. Human operators can approve or override agent actions; Shell intends to increase automation for routine cases while preserving human control for complex or high-risk decisions.

Stephen Ehikian, president of C3 AI, described the expected economic impact: “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.”

Microsoft, which provides Azure infrastructure for the deployment, supported the implementation. Sandy Gupta, a Microsoft vice president, noted: “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.”

Shell says the system will allow automation of the full maintenance lifecycle for certain categories of alerts, shortening the time between prediction and repair. The company aims to reduce unplanned downtime, avoid unnecessary preventive work on equipment that is operating normally, and align agent actions with existing ERP and planning systems.

The rollout builds on several years of using the C3 AI platform and forms part of Shell’s wider programme to improve reliability, safety and operational efficiency. As agents prove reliable for defined alert types, Shell plans to extend automated responses while keeping human review for non-routine situations.