Despite slow uptake, GenAI promises to be a disruptor in energy

Ever since the first public release of OpenAI’s ChatGPT model in late 2022, generative AI has been reshaping businesses and driving innovation across industries. At Enverus’ EVOLVE conference May 12-15, the “Generative Edge: Visionary Leadership in Energy’s AI Renaissance” panel delved into how the technology is being applied across the energy sector. Moderated by Enverus senior director Akash Sharma, the panel included Devon Energy chief technology officer Trey Lowe, Amazon Web Services products and solutions director Ben Wilson, Nvidia developer relations director for energy Srikanth Kodali and Enverus chief innovation officer Colin Westmoreland.

Wilson said implementations to date in the energy sector are small relative to other industries like media and entertainment, highlighting specific use cases such as using agents to evaluate well data to determine if a workover is needed or not. Kodali echoed that sentiment, saying Nvidia’s energy customers move more slowly—owing to energy’s highly regulated environment—but there is interest across the value chain. He added that customers are interested in using AI in safety, reliability, renewables integration, data mining and optimizing trading and logistics.

Lowe said Devon is differentially investing in GenAI relative to other oil and gas companies, noting that the company aims to save $250 million through production optimization—built largely on scaling machine learning models. During Devon’s Q1 call, Lowe said the company has invested in standardizing sensors across its thousands of wells and is working to use that data in real time with physics models and algorithms to determine optimal flowing conditions for each well.

Leaning into GenAI is not an easy task, however, and requires a top-down cultural and philosophical shift, Westmoreland said. Across AWS’ hundreds of thousands of customers in its AI business, Wilson said the single biggest indicator for success in adopting new technologies is putting a profit-and-loss leader in charge and holding them accountable for the results. Devon is in the process of setting baseline expectations for all employees to know and use GenAI tools, Lowe said, and will also reward and recognize their use. Still, the tech sector’s mantra of “move fast and break things” may not be as applicable to the energy industry.

“When we go for a home run, it’s high risk. When someone brings a new technology—and GenAI is one of these today—and they say we want to go use GenAI to find oil in a new part of the world where it’s never been found or we want to use GenAI to introduce a laser bit so we can drill wells in two days, this is where ideas go to fail for us,” Lowe said. “So, we have to be way more methodical in picking our bets with this. …We try to hit a lot of singles and we try to get adoption at a broad scale. We do invest heavily in our data so if one of these singles turns out to work, immediately you can copy/paste across thousands of operations all at once.”

Although the energy industry may be moving more slowly to adopt GenAI than other sectors, it still promises to be a disruptor in the coming years. Wilson said the biggest disruption he sees in the future is autonomy of control systems. He noted that a sand plant south of San Antonio integrated GenAI to run its control system, resulting in a 6% reduction in the use of gas to heat the sand and an 8% increase in sand production. AWS is also working with two companies, Wilson said, to automate entire power plants, which the companies believe will drive down costs and increase electricity generation.

The largest disruption Lowe has seen is in the pace of developing ideas. Previously, if an individual had an idea but didn’t know how to do it, they might get tasked to work on it for a month with a developer or data scientist. Now, Lowe said, the individual works with a large language model for an afternoon and they have a prototype.

“You can do that 1,000 times over the course of a week and all of a sudden you are moving at a pace you never have before,” Lowe said. “That is what’s happening today. We’re building all sorts of solutions and people are taking ideas to prototype in an hour or two hours, and this is something pace-wise that just never happened. And we’re starting to see our service providers and the companies we rely on doing the same, and that level of acceleration is exciting. It’s not robots and control systems yet, but you can definitely have a glimpse out on the horizon that that’s coming.”

In final takeaways, Lowe said data quality is foundational—if companies want to connect GenAI tools and do it at scale, they have to focus on their data. Kodali advised focusing on developing skills of employees and teams, as well as identifying one problem worth solving. Westmoreland wrapped up the panel saying GenAI adoption has to be top-down, adding, “if you feel your organization is observing, then you’re probably falling behind.”