Artificial intelligence (AI) offers vast potential, but tapping into that requires a change in mindset and a willingness to implement the technology.
There are other challenges as well, industry leaders said during the “AI for Energy, Energy for AI” strategic roundtable at CERAWeek by S&P Global on 11 March. AI is enormously power-hungry, which means energy demand will grow alongside uptake of AI technology. AI is only as good as its data, so quality and accessibility of data are both important. There’s also the question about the future of workers, given their worries about being replaced by technology.
Balaji Krishnamurthy, vice president of Chevron Technical Center, said AI can help predict well failures and identify emissions plumes and could help balance the future energy system. When considering AI solutions, he said, it’s important to keep the business case in mind and understand how it’s going to create value and change workflows.
Kraken CEO Amir Orad said AI can make everything around it more efficient.
“We also ask the question, ‘Are we saving more energy or using more energy by (using) the AI?’ ” he said.
Rakesh Jaggi, president for digital and integration at SLB, said he believes AI will accelerate exploration and discoveries, and Hitatchi Energy CEO Andreas Schierenbeck said he sees AI as having value for speeding up repetitive tasks.
Ben Wilson, director of products and solutions at Amazon Web Services, emphasized AI’s ability to provide context beyond mere information. For instance, he said, he may know routes to the airport but an AI-assisted navigation app provides an efficient route that avoids traffic jams. Similarly, he said, AI can provide context in industry operations. For example, when interrogating why a well’s production changed, there could be two answers of varying usefulness. One is that production changed because the choke was adjusted.
The industry already knows that, when the choke is adjusted, production will change, he said.
“But why did it change? Give me the context,” Wilson said.
The context could be that the well was prone to liquid loading and, based on previous data, it looked like the well was beginning to liquid load, he said.
“That’s the answer you’re looking for. That’s the answer AI can give you,” he said. “This is the way to kind of think about things in future.”
But AI is a power-hungry technology, and energy providers are trying to figure out how to meet anticipated demand.
Jaggi said, “There is no scenario where the power demand doesn’t grow.”
Lynda Clemmons, the senior vice president, chief sustainability officer, and head of strategy implementation at NRG Energy, said the expectation is that Texas alone will require the addition of the equivalent of California’s energy load to meet demand.
“That’s a lot of growth to wrap your head around,” she said.
Work Shift
While AI can make some operations autonomous, there’s an argument for keeping people involved.
Krishnamurthy said, “We want the human in the loop to make higher quality decisions.”
Wilson said he thinks of it as having a human in the lead.
And, increasingly, Jaggi said, the industry needs humans who have both domain expertise and data science knowledge.
But having the knowledge is only part of the equation. Having good—and accessible—data is critical.
Krishnamurthy said the industry collects more than a terabyte of data daily but people have spent as much as half of their time looking for the right data in the past.
Metadata helps make data findable, Wilson said.
“It’s not about how do I get everything 100% perfect, it’s how do I add more metadata so it can be more easily found,” he said. “I think we’re going to find that, for every one kilobyte of data we have, we have maybe two, three, four kilobytes of metadata describing that data, and that’s where real value and insights are going to come out.”
For years, the industry has talked about the pros and cons of standardizing data. AI may render the question moot, or the debate may continue.
Krishnamurthy said rules and standards are needed and noted getting legacy data into usable form is a challenge.
Jaggi said the entire industry must travel the standardization and transformation journey together.
Wilson agreed that some standards would help and pointed to the OSDU Data Platform as an example.
On the other hand, Schierenbeck said, it would take an enormous amount of time and effort to standardize legacy data. But AI could help by making sense of charts, pictures, handwriting, and other forms of legacy data, he said.
Orad said that, with AI breaking down data format issues, the bigger issue is around access.
“The silos are a much bigger issue than the format of the data,” he said. “You have a hodgepodge of 30 systems that don’t talk to each other, is a bigger challenge to solve.”
Another question that arises around using AI is how to do so while maintaining security.
Orad said there was a time when people were worried the cloud could compromise their operations, but, in fact, they have found it can be safer while speeding up innovation.
“Even the CIA [Central Intelligence Agency] is running on the cloud,” he said. “You can achieve security, but you have to change the way you think about things.”
Doing so can help companies tap into potential, he said.
Difficult Change
Trying to change the way things have been done can be met with resistance.
Schierenbeck said employees might worry about being made redundant because AI technology may do their jobs better than they do.
“They fear the change will affect their jobs. If you cannot take that fear away, they will find excuses about why this is not good, why it can’t be done, why it’s dangerous. It’s human nature,” he said.
Jaggi said some resistance may stem from people who are used to doing things a certain way. And, if a technology doesn’t work “perfectly out of the box,” they often turn into naysayers, he added.
Orad said people who tried AI a while back could have a very different experience if they try again now. A year ago, he said, AI was “like a child” and now it’s more accurate and efficient.
“Whatever you tried a year ago is irrelevant,” he said.
Schierenbeck said true support for using the tech must come from the top level of the company.
Wilson added that change management must revolve around keeping the human in the lead. For example, a human should be in the lead in a power plant to make final decisions on AI recommendations.
“Maybe someday it becomes fully autonomous, but that’s not what you want on Step 1,” he said, adding Step 1 might be leveraging the AI to change set points more frequently. “The action people really need to start thinking about is, ‘How do I take those first few steps?’ Because once you take those (steps), the next ones become more and more obvious.”
