ADNOC and OpenAI bring AI agents to energy operations

On November 2, 2026, ADNOC announced a multi-year partnership with OpenAI and G42 to deploy OpenAI Frontier in energy operations. The company says it is the first in the sector to use the platform to connect agents and decisions across the entire chain, from wells to market. (ADNOC announcement)

The news is not a new language model available to the public. It is a corporate deployment that combines GPT models from OpenAI, industrial data from ADNOC and sovereign AI infrastructure from G42 in the United Arab Emirates.

In this article you will see:

• What OpenAI Frontier will do at ADNOC
• The first two use cases announced
• How G42, agents and human oversight fit in
• What already existed in ADNOC's AI operations

What ADNOC announced with OpenAI and G42

The deal was announced in Abu Dhabi during ADIPEC and includes OpenAI Frontier, the platform the company describes as a way to coordinate AI agents in business operations. The proposal is to connect systems that previously supported separate decisions.

ADNOC already uses AI in subsurface and operations processes. The partnership aims to connect that intelligence to commercial factors.

The release describes a multi-year partnership, but it does not announce a public API, a new base model or a product that any developer can contract. The announced scope is the transformation of internal ADNOC processes, with solutions co-developed by the three companies.

The Abu Dhabi Media Office also recorded the meeting between Sam Altman and Crown Prince Khaled bin Mohamed bin Zayed and the signing of agreements with OpenAI. The photo below comes from that official record, not from a product demonstration. (Abu Dhabi Media Office record)

ADNOC integrates oil field planning

The first announced use is integrated field planning. Today, reservoir, facilities and economics studies can occur in separate stages. The proposal is to bring them together in a single decision on how to develop the entire field.

ADNOC says it will integrate decades of data from thousands of wells, several reservoirs and hundreds of surface facilities. Capital costs, operating costs and commercial factors would also be included to prepare development plans.

According to the announcement, the goal is to produce and optimize these plans in hours instead of months. That timeframe is an expectation published by the company for the new workflow, not an independent result already measured in production.

The intended gain is to plan the entire field, instead of optimizing each reservoir or facility in isolation. The source has not yet published an independent assessment of the case.

The partnership expands a trend of integrating enterprise AI into existing workflows. On the blog, the expansion of the partnership between OpenAI and ServiceNow shows another example of AI tools connected to corporate processes, although in a different sector and with a different scope.

ADNOC agents will link production, market and margin

The second announced use case aims to assess the value of each barrel while it is still underground. To do that, interconnected AI agents must link production, processing, refining, logistics, sales and commercial negotiation.

ADNOC says these agents will track supply, demand, price and margin forecasts, as well as estimate impacts before they happen. The intent is for the team to compare operational and commercial consequences in the same process, instead of consulting separate systems.

This does not authorize the system to control the entire operation. ADNOC says human specialists will supervise relevant decisions, but it has not yet detailed permissions or approvals.

For a complementary overview of agent limits, see how OpenAI describes misalignment monitoring. The ADNOC announcement does not say that Frontier eliminates failures or replaces existing safety controls.

How G42 and the human team fit into ADNOC's deployment

The solution will be developed jointly and run on G42's sovereign AI infrastructure in the United Arab Emirates. ADNOC says its proprietary industrial data and specialized knowledge will be combined with OpenAI's advanced models.

The release also states that the solutions will operate under ADNOC supervision, with protection for proprietary data and critical infrastructure. These commitments describe the announced architecture; they are not equivalent to an independent public audit of security or data residency for all data.

The partnership starts at a company that already uses AI at scale. ADNOC reports having more than 200 AI applications and 70 robotics use cases deployed or in development. These numbers contextualize the program, but they are not results produced by the new agreement.

ADNOC also says ChatGPT Enterprise has been made available to more than 40 thousand employees, with adoption above 97%. The release does not detail the method used to calculate adoption, the observed period or usage frequency, so these numbers should be read as data disclosed by the company itself.

Among existing systems is ENERGYai, focused on subsurface workflows. ADNOC says it uses a 70 billion parameter model trained with more than 50 years of company data. The new partnership is described as a layer that connects that intelligence to other decisions, not as a replacement for ENERGYai.

The company cites a real time operations center that connects more than 120 wells and reduced engineering effort by 30% to 40%. That result belongs to the current operations described by ADNOC, it is not a performance metric for the agreement with OpenAI.