Woodside Energy says it now has about 50 AI agents running in production across operating assets and enterprise workflows, an expansion that shows where some companies are putting AI to work outside the usual chatbot circus.
Andrew Melouney, Woodside’s vice president for digital, told MIT Technology Review’s Business Lab that the Australian energy producer has been using analytics, optimization systems and machine-learning models since around 2015. The company’s AI work started with industrial data, he said, because Woodside runs large physical assets that generate continuous streams of sensor and equipment information.
That distinction matters. In an LNG plant, a bad suggestion is not an awkward email draft. Woodside’s use cases sit inside maintenance planning, plant operations, drilling, exploration and trading workflows, where safety, reliability and uptime are part of the job rather than marketing garnish.
From predictive maintenance to AI agents
Melouney said Woodside built its current AI program on an enterprise data platform that ingests high-frequency operational data from assets and information from business systems. He described that governed data layer as the reason the company can build models that operators might actually trust.
One example is a maintenance intelligence system. According to Melouney, Woodside combines historical maintenance records, including data from SAP, with equipment performance data from its time-series data lake. The system analyzes those data sets together and recommends when maintenance should happen.
Melouney said a pilot on one asset showed an opportunity to cut maintenance hours by up to 15% over five years. That is Woodside’s estimate, not an independently verified result. The point of the system, as he described it, is decision support: asset teams remain responsible for deciding what work gets done.
Woodside is also applying agent-based systems to LNG plant startups. Its Startup Advisor is designed to sit beside control-room operators during the complicated process of bringing an LNG facility online. Melouney said it can show prior startups, track how the current startup is progressing and provide information intended to help operators run the process more consistently.
The company’s stated design principle is augmentation rather than replacement. Melouney said Woodside wants AI tools to help employees make decisions faster and with better information, while keeping accountability with human operators.
Governance before scale, at least in theory
Woodside’s pitch is that industrial AI scale depends less on demos and more on boring infrastructure: standardized platforms, repeatable build patterns and rules for deployment. Melouney said the company does not want 50 tools built 50 different ways.
He said each AI use case goes through a structured review covering privacy and cybersecurity controls. If concerns remain, the proposal goes to an AI council made up of senior leaders who review risk, prioritization and whether the company should deploy the system at all.
Woodside is also trying to manage agents after launch. Melouney said the company tracks usage, effectiveness, model drift and whether systems need retuning or retraining. He acknowledged that monitoring 50 agents is easier than managing hundreds or thousands.
Infosys also sits inside the story. The Business Lab episode was produced in partnership with Infosys, and Melouney said Infosys is Woodside’s managed service provider for core systems. He said the partnership helps keep basic platforms running and gives Woodside access to additional AI talent through mixed teams, while Woodside keeps ownership of strategy, governance and outcomes.
Melouney described Woodside’s longer-term goal as an “autonomous enterprise,” with agents able to interact with core workflows across exploration, projects, operations and marketing. For now, the confirmed facts are narrower: Woodside says it has production AI agents, governed operational data and specific systems aimed at maintenance and LNG startup support. The autonomy part remains an ambition.
This story draws on original reporting from MIT Technology Review.
