AI Implementation

AI in UAE Oil and Gas: How It Actually Works

A sourced look at the AI actually running in UAE oil and gas today, from ADNOC's Panorama command centre and the 340 million dollar ENERGYai agentic rollout to the SLB drilling operations centre on 120 rigs, and what the sequence means for smaller firms.

September 15, 202610 min readBearingNorthAI

Industrial oil and gas processing facility with extensive red and white pipework, representing UAE upstream operations now running on AI systems — Photo by Christian Harb on Unsplash Source

The UAE's oil and gas sector is the most AI-heavy part of the country's infrastructure economy, and it is not close. While most sectors are still arguing about pilots, ADNOC signed a 340 million dollar, three year contract in March 2025 to deploy agentic AI across more than 28 producing fields, and by August 2026 had an AI drilling platform running on over 120 rigs. That is not a roadmap. That is production.

This guide covers what is actually running inside UAE oil and gas today, with dates and numbers attached: the command centre layer ADNOC built first, the subsurface and maintenance systems that followed, the ENERGYai agentic platform now scaling across upstream assets, and the drilling operations centre delivered with SLB. It then covers the part that matters more to most readers, which is what any of this means if you are a service company, contractor, or mid-sized operator rather than a national oil company.

Why Oil and Gas Got There First

The sector had three advantages that most UAE industries do not. It had decades of structured operational data already digitised, it had capital budgets large enough to absorb a failed experiment, and it had problems expensive enough that a small accuracy gain pays for the whole programme. A single day of unplanned rig downtime costs more than most SMB AI projects cost in a year.

The International Energy Agency made the same point at a global level in its Energy and AI report, published 10 April 2025, which identifies oil and gas as an early adopter of AI and maps the use cases to exploration, production, maintenance, and safety: evaluating resources more reliably, reducing drilling uncertainty, optimising production, detecting leaks, predicting maintenance needs, and supporting methane reduction. The UAE did not invent this list. It funded it harder and earlier than most.

The Foundation Layer: Panorama and the Command Centre Model

Before any of the current AI work, ADNOC built a data spine. The Panorama Digital Command Center aggregates real-time information across ADNOC's 14 subsidiary and joint venture companies and applies analytical models, AI, and big data to produce operational recommendations. By May 2020 ADNOC reported Panorama had generated more than 1 billion dollars in value (AED 3.67 billion) since it went live, and it became the mechanism for running the business remotely through COVID-19.

This sequencing is the single most transferable lesson in the whole story. ADNOC did not start with an AI model. It started by getting operational data from 14 separate entities into one place that a model could later read. Firms that skip this step end up with the legacy integration problem instead, which is a slower and more expensive place to discover it.

The Tool Layer: 30 Systems and 500 Million Dollars

By March 2024, ADNOC had deployed over 30 AI tools across its value chain and reported 500 million dollars (AED 1.84 billion) in value generated in 2023 alone, alongside up to 1 million tonnes of CO2 abated between 2022 and 2023. The named systems give a useful picture of where the value sat:

Thamama, a subsurface centre of excellence applying AI to reservoir and geological work
CPAD, a centralised predictive analytics and diagnostics system for equipment health
Emission X, used for emissions detection and reduction
SMARTi, a computer vision system
AR360, for reservoir visualisation
Robowell, for remote operation of wellhead equipment

Notice how much of that is maintenance and equipment health rather than anything exotic. The same pattern shows up in utilities, which we covered in detail in our guide to AI predictive maintenance for UAE utilities. Predictive maintenance is boring, measurable, and usually the first thing that pays for itself in an asset-heavy business.

The Agentic Layer: ENERGYai

The current phase is different in kind, not just in scale. ENERGYai, developed by AIQ with G42 and Microsoft, is built as an agentic system: AI agents that carry out multi-step tasks and return findings in natural language rather than dashboards a human has to interpret.

The technical shape, from ADNOC's announcement on 16 January 2025, is a 70 billion parameter large language model trained on over 50 years of ADNOC operational knowledge and petabytes of proprietary data. The 90 day proof of concept drew on data from more than 15 percent of ADNOC's onshore and offshore wells and reported a 70 percent improvement in accuracy on major seismic interpretation tasks, plus gains in reservoir monitoring and anomaly detection. Over 100 specialists worked on it.

Two months later, on 12 March 2025, AIQ announced a 340 million dollar, three year contract to deploy ENERGYai and related solutions across ADNOC upstream operations covering more than 28 producing fields, starting with five fully operational agents for subsurface tasks and scaling to thousands of additional wells. The stated benefit is compression of process time from months to days.

It is worth being precise about what that 70 percent figure is and is not. It is a proof of concept accuracy improvement on specific seismic interpretation tasks, measured by the parties deploying it. It is not an independently audited productivity number for the business as a whole, and nobody should quote it as one. The discipline of separating vendor-reported pilot metrics from verified operational impact is the core of our framework for measuring ROI from AI implementation.

The Drilling Layer: ADNOC and SLB's Real-Time Operations Center

The most concrete recent deployment is also the most useful one to study, because the numbers are operational rather than financial. On 4 August 2026, ADNOC and SLB announced the deployment of a Real-Time Operations Center across over 120 rigs in ADNOC's fleet. The platform analyses drilling data as it arrives, flags risks before they escalate, and turns raw telemetry into automated dashboards and analytics.

The reported results, per that announcement:

Engineering effort reduced by 30 to 40 percent
Each engineer able to oversee two to three times more rigs
Reporting cycles cut from several days to hours
Incident response times reduced by 4 to 12 hours
One to two days of rig downtime prevented

Musabbeh Al Kaabi, ADNOC Upstream CEO, and Rakesh Jaggi, President of Digital at SLB, both framed it as a performance and safety play rather than a headcount one. The honest read of those numbers is that the value is in supervision span. The same engineer covers more assets and sees problems earlier. That is what AI is genuinely good at in heavy industry today, and it is a more defensible business case than any claim about replacing people.

What Changed in 2026: Pilots Became Infrastructure

The strategic shift is the part most UAE firms should pay attention to. Computer Weekly reported on 4 August 2026 that ADNOC has repositioned AI as an enterprise capability rather than a portfolio of individual applications, moving away from standalone digital initiatives and embedding AI into everyday workflows across exploration, drilling, production, maintenance, and decision-making. ADNOC says its internal AI Lab approach cut the time from proof of concept to deployment by a factor of three.

That factor of three is arguably the most important number in this article. Most AI programmes do not fail at the model. They fail in the gap between a pilot that worked and an operation that uses it, which is exactly the transition we mapped in our guide on moving from AI pilot to full rollout. ADNOC's answer was structural: a standing internal capability whose job is to shorten that gap, rather than treating each deployment as a fresh project.

The commercial consequence is already visible. In June 2026, AIQ CEO Dennis Jol told Semafor that the company had built 200 AI use cases at ADNOC and was now exporting the products, targeting the US first, then the North Sea, Canadian oil sands, and the global south. The UAE is no longer only buying energy AI. It is selling it.

What This Means If You Are Not a National Oil Company

Most UAE businesses reading this are somewhere in the supply chain rather than at the top of it: a drilling services firm, an inspection contractor, a fabrication yard, a logistics provider, an engineering consultancy. Three practical implications follow.

First, the data expectations travel downstream. When an operator runs real-time analytics across its rigs, the contractors feeding data into that environment are expected to deliver it in a usable, timely, structured form. Firms still sending scanned PDFs and end-of-week spreadsheets become the friction point in someone else's automated workflow, and that is a commercial risk long before it is a technical one.

Second, the highest-value AI work for a smaller firm is almost never at the frontier. It is document and knowledge work: finding the right specification across ten years of project files, assembling a tender response from previous submissions, extracting data from supplier documents, answering the same technical enquiry for the fortieth time. These are retrieval problems, they run on the documents a firm already owns, and they do not require a 70 billion parameter model or a petabyte of telemetry.

Third, the vendor landscape has consolidated around a recognisable set of names. AIQ, G42, Presight, Core42, and the global hyperscalers behind them now cover most serious UAE energy AI work, which we break down in our guide to the top AI providers for UAE infrastructure firms. Knowing who actually builds what saves a lot of time in procurement.

A Realistic Starting Sequence

If you run a mid-sized firm in or around this sector, the ADNOC story compresses into four steps that scale down surprisingly well:

Consolidate first. Get operational and document data out of individual inboxes, shared drives, and departmental spreadsheets into one place a system can read. This is unglamorous and it is the step everyone skips.
Pick a process with a measurable clock. Time to produce a report, time to answer a technical enquiry, time to close out an inspection. If you cannot state today's number, you cannot prove tomorrow's improvement.
Run a bounded pilot with a baseline recorded before go-live, not reconstructed afterwards.
Build the deployment path before you need it. Decide in advance who owns the tool, who maintains it, and how it reaches daily workflows. This is ADNOC's factor of three, available to a 60 person company at a fraction of the cost.

The Honest Summary

UAE oil and gas is genuinely ahead on AI, and the evidence is dated, named, and specific rather than aspirational. The caveat is that nearly all the published performance figures come from the organisations deploying the systems, and the financial value numbers are internally calculated. That does not make them wrong. It does mean a smaller firm should treat them as directional evidence that these use cases work, not as benchmarks to promise a board.

The transferable lesson is not the technology. It is the sequence: data foundation, narrow measurable use cases, then a standing capability whose job is getting things from pilot into daily operation. That order held for a company with 28 producing fields, and it holds for a company with 28 employees.

Research Sources Used

FAQ

Common questions.

Which AI systems does ADNOC actually use today?

ADNOC has reported more than 30 AI tools across its value chain, including the Panorama Digital Command Center for enterprise-wide operational data, Thamama for subsurface work, CPAD for predictive analytics and equipment diagnostics, Emission X for emissions, SMARTi for computer vision, AR360 for reservoir visualisation, and Robowell for remote wellhead operation. Its current flagship is ENERGYai, an agentic AI platform built by AIQ with G42 and Microsoft, plus a Real-Time Operations Center deployed with SLB across over 120 drilling rigs.

What is ENERGYai and how large is the deployment?

ENERGYai is an agentic AI solution for the energy sector built by AIQ. ADNOC describes it as a 70 billion parameter large language model trained on over 50 years of ADNOC operational knowledge. In March 2025, AIQ announced a 340 million dollar, three year contract to deploy it across ADNOC upstream operations covering more than 28 producing fields, beginning with five operational agents for subsurface tasks and scaling to thousands of additional wells.

Are the reported AI results independently verified?

Mostly no. The accuracy improvements, value figures, and efficiency gains cited in this article come from ADNOC, AIQ, and SLB announcements rather than independent audits. They are credible as evidence that these use cases are in production at scale, but they are company-reported metrics and should not be used as guaranteed benchmarks for another organisation's business case.

Can a smaller UAE energy services company use any of this?

Yes, though not by copying the technology. The transferable parts are the sequence and the use case selection. Consolidating operational and document data into one accessible place, choosing a process with a measurable time baseline, and deciding in advance how a successful pilot reaches daily workflows all apply at any size. For most smaller firms, the highest-return applications are document and knowledge retrieval rather than subsurface modelling.

Why is the UAE ahead of most markets on energy AI?

Three structural reasons: decades of already-digitised operational data, capital budgets large enough to absorb failed experiments, and problems expensive enough that small accuracy gains justify large programmes. National strategy and sovereign compute investment through G42 and its affiliates accelerated it further, to the point that AIQ is now exporting energy AI products built on ADNOC deployments to the US and other markets.