Woodside Energy is embedding agentic AI into the operation of its liquefied natural gas plants, layering autonomous agents over a data and machine learning foundation the company has been building since around 2015. The Western Australia-based producer is now targeting what its digital chief calls an autonomous enterprise, in which AI agents plug directly into core industrial workflows across exploration, drilling, maintenance, and plant operations. The work is being done in partnership with Infosys, which co-produced the July 2, 2026 Business Lab episode where Woodside detailed its approach.
Andrew Melouney, Woodside's vice president for digital, said the company has spent more than a decade collecting and governing operational telemetry from remote assets, sensors, and production equipment. That long-run investment is what makes today's agentic layer feasible, he argued — the models sit on top of a curated, enterprise-scale data platform rather than on ad-hoc extracts.
“We've always had very large volumes of operational data coming from the equipment and the plants and the assets that we operate”— Andrew Melouney, Vice President for Digital, Woodside Energy
The flagship deployment is a copilot called Startup Advisor, which guides operators through the multi-step process of bringing an LNG plant online. LNG startups are safety-critical and procedurally dense, and errors are expensive. Rather than automate the operator out of the loop, Woodside built the system to compress decision time and surface the right information at the right step.
Key facts
- 01Woodside Energy has been applying machine learning and predictive analytics across its operations since around 2015, more than a decade before its current agentic push.
- 02The company's Startup Advisor is an AI copilot that guides operators through the process of starting LNG plants in Western Australia.
- 03Andrew Melouney, VP for Digital, says Woodside's ambition is an autonomous enterprise where agents interact directly with core workflows.
- 04The initiative is being developed in partnership with Infosys and detailed on the July 2, 2026 episode of MIT Technology Review's Business Lab.
Melouney frames the design philosophy as augmentation, not replacement. The goal, he said, is to support people in the organization so they can make better decisions and faster decisions in environments where the physical stakes — equipment, personnel, remote assets — are real. That framing distinguishes industrial AI from the consumer chatbot boom that has dominated headlines.
The company is also being explicit that this is not a bolt-on exercise. Woodside is redesigning the underlying workflows in parallel with the AI deployment, on the theory that grafting a model onto an unchanged process yields marginal gains at best. Melouney's operating motto for the program is: think big, prototype small, and scale fast.
Industrial AI has historically lagged consumer AI in visibility but often leads in dollar impact. Reliability improvements in an LNG train, incremental gains in predictive maintenance, and faster plant startups translate directly into throughput and revenue. Woodside's earlier machine-learning work targeted exactly those levers — analytics, optimization, and predictive models across a global asset base — and the agentic layer is being built on the same value logic.
The next phase, Melouney said, is agents with genuine agency inside core workflows: systems that don't just recommend but act, within bounded and governed contexts. That raises the governance bar sharply, which is why Woodside emphasizes data quality, human accountability, and trust between the digital organization and the operational business as prerequisites rather than afterthoughts.
“Our ambition is really for an autonomous enterprise, where we have agents with agency that are able to really deeply interact with our core workflows”— Andrew Melouney, Vice President for Digital, Woodside Energy
Skeptics of industrial agentic AI point to the gap between demoware and production reliability in safety-critical environments. An agent that hallucinates in a chatbot is annoying; an agent that misreads a sensor feed during an LNG startup is a hazard. Woodside's answer is the decade of data governance work underneath the models, plus a deliberate scoping of where agents get to act versus where they only advise — but the company will need to show sustained reliability across many startups before the pattern is proven.
For the broader AI market, Woodside's playbook is a useful counterpoint to the consumer-app narrative. The companies most likely to extract durable value from agentic AI are the ones that spent years cleaning and governing the operational data that agents need to act on. That is a moat that OpenAI, Anthropic, and Google cannot sell — it has to be built inside the operator. Expect more asset-heavy industries to follow this template, and expect the systems integrators wiring it together, Infosys among them, to capture a meaningful slice of the industrial AI budget over the next several years.
Working on something we should cover, or seeing a story we missed? Send leads, documents, or feedback to hello@aichatdaily.com. For sensitive tips, see our secure tips page for Signal and PGP options.
Spotted an error? Email hello@aichatdaily.com with the URL and the issue, or read our full corrections policy.
