ADNOC goes all-in on agentic artificial intelligence. Enverus draws on generative AI. Neural networks do battle – with each other. And the cyber menu expands for the oil and gas industry, as energy companies cook up new recipes for digital exploration.
Agentic AI
In November, the Abu Dhabi National Oil Company launched ENERGYai, calling it “the world’s first-of-its-kind, custom-built agentic artificial intelligence solution for the global energy transformation.”
Agentic AI gets its name from the creation of programmed “agents” that operate at a near-human cognitive level. In many cases, those AI agents can take action autonomously – that is, with minimal human decision-making or intervention involved.
The new tool and platform “marks a major milestone in ADNOC’s journey to be the world’s most AI-enabled energy company and a key enabler of the global energy transformation,” said Sultan Ahmed Al Jaber, ADNOC CEO and minister of industry and advanced technology for the United Arab Emirates.
“As AI accelerates the pace of progress and transforms industries, ENERGYai will help future-proof ADNOC, reinforce our position at the forefront of AI deployment and ensure we continue to provide secure and sustainable energy to the world,” Al Jaber said.
Generative AI
Earlier this year, Enverus introduced its trademarked “Instant Analyst” technology using generative AI. With the new analysis tool, users can directly query a dataset containing multiple types of source documents and information, including proprietary research.
“The energy industry is experiencing a fundamental transformation through AI, particularly in exploration,” said Akash Sharma, product director for Enverus.
“Traditional data analysis methods are being augmented by sophisticated AI systems that can process and interpret vast amounts of geophysical and geological data at unprecedented speeds,” he said.
Generative AI recognizes underlying structures and patterns in data, goes through an analysis process and then generates original output – in images, text or other forms of information. Output can be triggered in response to a specific prompt or query.
“The emergence of generative AI represents a significant leap forward. Its advanced pattern-recognition capabilities are particularly valuable in understanding complex subsurface characteristics,” Sharma noted.
Enverus is a software-as-a-service (SaaS) data, information and analysis platform dedicated to the energy industry. Sharma said the company has seen AI revolutionize exploration workflows in oil and gas.
“Our research shows that generative AI’s ability to rapidly analyze and correlate massive datasets is helping exploration teams uncover insights that might have taken months to discover through conventional methods,” he said.
“We’ve learned that successful AI implementation requires more than just technical expertise – it demands a deep understanding of energy industry workflows and challenges,” Sharma added.
Evolution Driven by Data-Processing Demand
High-performance computing and AI began to transform exploration decades ago, particularly utilizing 3-D seismic, said Fred Aminzadeh.
Aminzadeh is a former president of the Society of Exploration Geophysicists and co-author of a 2022 book on artificial intelligence in exploration and production. He’s currently working on a book about AI for carbon capture and storage.
“We knew how to do it (effectively apply 3-D seismic to exploration), but we didn’t have enough computer power to crunch the numbers,” Aminzadeh recalled.
“Later on, we also got into 4-D seismic, which is a little further away from exploration – 4-D seismic involves more data, an order of magnitude larger,” he said.
AI has been especially helpful in combining and analyzing very large and very disparate data sets, Aminzadeh said. He uses the acronym SURE to describe the challenge of bringing together different data types.
Aminzadeh derived its name from (1) the Scale of data, (2) the Uncertainty level of data, (3) the Resolution of data and (4) the Environment where the data is gathered.
“One main application of AI to exploration is integrating different types of information,” Aminzadeh observed. “We are trying to combine all these data types. AI has been useful, and can be even more useful, in integrating all kinds of data involving the SURE challenge.”
Sharma sees the ability to manage data differences and data complexity as one of the primary challenges in applying AI tools in energy. He listed it first among several considerations:
- Data complexity: Historical energy data often exists in various formats and quality levels, requiring sophisticated normalization.
- Trust and verification: Given the high stakes in exploration, AI systems must provide transparent reasoning behind their recommendations.
- Industry-specific nuances: Every basin and formation has unique characteristics that require specialized AI models.
- Integration complexity: Incorporating AI into established workflows requires careful consideration of existing processes.
Oil and gas information also can be divided into the distinct categories of publicly available data and sometimes closely held proprietary data and analysis. At Enverus, “a cornerstone of our approach has been the strategic integration of public and proprietary datasets, a crucial differentiator in the energy sector,” Sharma said.
“This unique blend of data sources, when combined with advanced AI algorithms, provides a more complete and nuanced view of exploration opportunities than either dataset could offer alone,” he noted.
Other Combinations, Iterations, Applications of AI
Recent developments in generative AI involve pairs of neural networks in competition as well as diffusion models that add noise to data sets and then reverse the process.
In a generative adversarial network (GAN), two neural networks combine to produce increasingly valid new data from a training dataset. One network modifies data from an input sample; the other tries to determine if the new output belongs in the original dataset, or if it is “fake.”
The first network generates data until the predicting network cannot distinguish modified data from the original data.
A diffusion model modifies training data by introducing randomness to the data, in the form of Gaussian noise. The model then learns to remove the noise and replicate the original data or to generate new data with the same parameters.
These deep-learning methods of generating data output have led to significant advances in imaging and hold implications for many other applications of AI.
“While we’re still in the early stages of AI adoption in E&P, the results are promising. Companies leveraging advanced AI solutions are seeing significant improvements in all aspects of exploration and development operations,” Sharma said.
“Generative AI is proving particularly transformative, fundamentally changing how exploration teams approach complex analyses and problems,” he noted.
Today, energy companies of all sizes are using AI tools to meld and analyze the huge datasets of information available to the industry. New applications are emerging in generative AI, agentic AI and other areas.
Aramco might have won the coolest name contest with its “Metabrain” generative AI model, introduced in March.
So far, “the key to success has been focusing on practical applications that deliver immediate value while building toward more ambitious long-term goals,” Sharma observed.
“Our collaboration with leading energy companies has shown that when properly implemented, AI can significantly enhance human expertise rather than replace it,” he said.