Energy: AI agents reach oil planning; utilities keep people deciding
Energy companies began giving AI agents planning and commercial work in the past six weeks, beyond the maintenance and inspection jobs AI has done for years. On November 2, ADNOC, Abu Dhabi's national oil company, signed a deal to run OpenAI's agent platform on field planning and on valuing its oil before it is produced. Siemens Energy agreed to build autonomous plant software with an ADNOC venture. Arizona utilities told their regulator that a person still makes the decision. The tools are reaching bigger decisions faster than anyone is publishing results.
Agents reach the oil planning desk
The biggest announcements since our last look came from ADIPEC, the oil and gas industry's annual conference in Abu Dhabi, held November 2 to 5. ADNOC said it will be the first energy company to use Frontier, OpenAI's platform for running AI agents: programs that carry out a multi-step task on their own. The system runs inside the United Arab Emirates on servers operated by G42, an Emirati AI firm. The first job is field planning: combining the studies of the rock underground, the surface facilities and the economics, now done one after another, into a single plan. The second is putting a value on each barrel while it is still underground. Both sides say people keep the consequential decisions. ADNOC expects planning to drop from months to hours, its own estimate; no price, model or measured result was published. Reuters and Forbes reported the deal.
AIQ is an ADNOC joint venture with an Abu Dhabi technology firm. At ADIPEC it signed with Siemens Energy to build an "autonomous operations" offering for energy and process plants, and the announcement names AIQ's RoboWell among its first products. RoboWell, co-developed with the US oilfield-services company Halliburton, runs a well's controls on its own, including gas lift (gas pumped into a well to help the oil rise). AIQ's technology chief told the Gulf trade outlet Arabian Reseller on November 7 that RoboWell now runs on more than 500 wells, up from 10 wells when it reached ADNOC's offshore NASR field in 2024. Those are AIQ's own figures, and no price, launch date or customer outside ADNOC has been published.
In the US, Constellation, the country's largest operator of nuclear plants, announced on October 6 a five-year alliance to build agents on Google's AI tools. According to the trade magazine POWER, the uses include preparing permit applications, planning grid connections and scheduling the maintenance shutdowns of its reactors.
Arizona utilities tell their regulator a person decides
On October 21, the Arizona Corporation Commission, the state's utility regulator, held a public workshop on AI in utility operations, part of an inquiry it opened in March. Presenters included Arizona Public Service (APS) and Salt River Project, the state's two largest electric utilities, and a rural electric cooperative. The commission's summary, published October 27, says the two large utilities stressed "human guidance and decision-making" in how they use AI. Nothing was voted on and no rule followed.
In New York, the first inventories of every AI system the state's utilities use fall due around November 16, under the order covered in Energy: New York orders utilities to list every AI tool they use. None had been published at the time of writing.
Small buyers get the tools through groups
The Florida Municipal Power Agency, which buys power and services for 32 city-owned utilities, made Noteworthy AI's grid inspection system available to all of them after a competitive tender, according to an October 15 release from the vendor. Cameras on ordinary utility trucks photograph poles and flag defects. It is the second US buying group to offer the system, after the Ohio-based American Municipal Power.
The Southern Maryland Electric Cooperative, a US rural utility, described a model it built in-house that predicts equipment failures within a three-day window and is retrained every two weeks, in a story published on October 13 by the US cooperatives' national lender. The cooperative says it chose this over building a chatbot for its members.
In Brazil, Neoenergia, the local arm of Spain's Iberdrola, says its distributor in the state of Rio Grande do Norte runs 13 AI-based tools to prepare for a possible El Niño, a Pacific warming pattern that brings storms and heat. One finds a fault, cuts off the damaged section and reroutes power, often with no one stepping in. Brazilian regional papers reported it on October 13, from the company's release.
What it means for you
The ADIPEC deals put agents on planning and commercial decisions at the scale of a national oil company. A small operator is unlikely to buy Frontier directly; the same pitch will reach it through vendors it already pays, and Siemens Energy is now building autonomous operations to offer them. Utility Dive, the US trade publication, reported in late October that AI agents have barely reached anything that touches physical equipment at US utilities, and that evidence of steady performance at scale is thin. Expect the sales calls before the proof.
Arizona gives you the sentence to use with your own regulator, lender or insurer: the tool recommends, a named person decides. Two large utilities said so in public before any rule asked them to.
For a municipal utility or cooperative, the buying group is now the cheapest route to AI inspection, and a narrow failure forecast, retrained often, is work a small analytics team can copy.
What to do about it
If you run wells and a vendor offers autonomous control, ask for results from a field outside ADNOC, measured by someone other than the vendor, and for the minimum number of wells in the contract. Until you have that, watch and wait.
If you answer to a state commission in the US or an energy regulator in Europe or Latin America, add one line to your AI policy before anyone asks: which tools may recommend, which may act, and who signs off. Then ask your buying group what AI inspection it already offers before you tender on your own.
- 28-Sep-2026Energy: New York orders utilities to list every AI tool they use
- 02-Sep-2026Energy: AI reaches the control room as the EU delays its deadline
- 21-Jul-2026Energy: the power grid fell short, and the backlash began
- 09-Jun-2026Energy: AI became the biggest new buyer of electricity