Brian Spector has a blunt message for his students about artificial intelligence, and it probably works just as well for energy industry professionals:
“Your worth to your future employer is you plus AI.”
Spector had a 30-year career in energy, left BP in 2015 and later launched an AI company focused on back-office work. He’s currently an adjunct professor of business and an adjunct professor in environmental studies at Rice University in Houston.
“People right now are feeling several different things about AI: A, they’re kind of scared. B, they think their job is at risk. C, they think they’re behind their peers and colleagues,” he said.
Those concerns aren’t unjustified.
“People are going to have to take personal responsibility in learning about these (AI) tools. You have to lean in. You aren’t going to be able to lean back,” Spector said.
Spector will be a speaker at the lunch session “How AI is Reshaping Subsurface Energy Systems” on Aug. 19 at the 2026 International Meeting for Applied Geoscience and Energy in Houston. AAPG’s Division of Professional Affairs is the luncheon sponsor.
He’ll be joined on the speaker’s platform by Sahar Bakhshian, assistant professor in earth, environmental and planetary sciences at Rice. She said her part of the presentation will focus on how AI transforms our understanding of subsurface systems, and what that means for the future of the energy industry.
“I will discuss the evolution from traditional modeling to data-driven and physics-informed AI approaches, and how these technologies are accelerating simulation, forecasting, monitoring and decision-making,” Bakhshian said.
She also plans to look at “emerging concepts such as digital twins, agentic AI and human-AI collaboration.
“Rather than viewing AI as a replacement for geoscientists and engineers, I see it as a powerful partner,” she said.
Both Spector and Bakhshian emphasized the importance of energy professionals understanding what AI can do for the industry and knowing how to use AI in their everyday jobs.
“For the attendees, I want them to feel empowered to use AI tools and use them in the right way. We want to be very clear on what AI is and what it does,” Spector said.
“I don’t believe every energy professional needs to become an AI expert or develop machine-learning models. However, they do need to understand what AI can and cannot do, where it adds value, and how to use it effectively in their workflows,” Bakhshian noted.
AI Taxonomy
“What is increasingly important is understanding how to ask the right questions, interpret AI-generated results, recognize limitations and integrate AI into decision-making,” she said.
Different types of AI applications have emerged for different purposes. Generative AI creates original output. Agentic AI can perform tasks, from basic to complex.
“What adds the most value in the upstream space is ‘deterministic AI,’” said Spector. “The benefit of deterministic AI is that it is solely trained on the data involved in your tasks, thus each question will yield the same answer every time, and you won’t get what’s called ‘hallucinations.’”
“As opposed to automation, AI’s predecessor, deterministic AI learns through experience, continually improving once given direction,” he added.
In that view, deterministic AI produces exactly the same results from the same input every single time, without randomness. And it can be developed and trained for specific needs in any given industry.
“The way you train a deterministic AI model is the same way you would train employees. It’s not that complicated. Using AI to handle the mundane, repetitive tasks enables employees to focus their time on pieces of the workflow that need human intervention,” Spector commented.
“Ultimately, what properly designed and employed AI models will do is let people focus on the highest-level tasks,” he said.
Spector noted that energy industry professionals don’t need to be concerned about how AI is programmed or worry about learning a special language for AI queries and commands.
“That happens in English, and is editable,” he said.
In subsurface analysis, AI can help throughout the entire subsurface workflow, Bakhshian said. For characterization, “AI can accelerate interpretation of seismic data, well logs, core images and other geological information. Tasks that previously required extensive manual effort can now be performed much more efficiently,” she said.
She talked about runtimes shrinking from days or hours to seconds in reservoir modeling and computationally intensive simulations.
“When we have sparse data, AI helps with prediction of missing logs, spatial interpolation between wells, multimodal data fusion – combining well logs and seismic, production history,” she said.
“Even though each dataset is sparse by itself, we can extract richer information if we combine them together, integrating multiple incomplete datasets and extract patterns that are not apparent in each dataset independently,” she added.
Synthetic data has become increasingly important in subsurface energy applications because of the limited availability of high-quality, labeled data, Bakhshian noted. She said it will play a critical role in the future of subsurface AI, providing a bridge between physics-based understanding and machine learning.
A combination of physics, synthetic data and limited field observations can be more powerful than relying on real-world data alone, she said.
“We often do not have millions of examples of reservoir behavior, leakage events, drilling outcomes, or subsurface processes. In many cases, collecting real-world data is expensive, time-consuming or impossible,” Bakhshian observed.
“Synthetic data helps address this challenge by generating realistic datasets using physics-based simulations, digital rock models, reservoir simulators, geomechanical models or other computational tools,” she said.
The Best Way to Bake Some Cookies
The oil and gas industry is just beginning to use the AI and physics approach to subsurface energy, according to Bakhshian, who added, “I think the industry is not very advanced. When it comes to physics-based AI it’s at the very earliest stage.”
Spector sees this in part as simply the result of the newness and unfamiliarity of AI applications.
“I think every industry is behind. The energy industry is always going to be a bit behind, because the industry is mostly made up of very large companies, and very large companies tend to be slower to change,” he noted.
“But relative to other industries, I don’t think the energy industry is measurably more behind,” he said.
Energy professionals should “recognize that AI is not replacing domain expertise,” Bakhshian said. “In fact, geological understanding, engineering judgment and physical intuition become even more important because humans must evaluate and supervise AI-generated recommendations.
“The future is likely to be one of human-AI collaboration rather than human replacement,” she said.
But some energy jobs will disappear as AI is integrated into industry operations, Spector said. He foresees those jobs being replaced by new types of jobs.
“In my opinion, work processes are going to change across all industries. And, yes, humans will be in the loop. But will they be the same humans, or will they be ones trained to employ these new tools?” he said.
He emphasized the importance of participatory training in teaching professionals how to use AI, and compared the process to learning how to make cookies. People can be trained to make cookies by researching cookies or studying cookie recipes or even watching cookies made.
However, the best training would involve actually making cookies.
“When I say ‘lean in,’ I don’t mean just naturally attend a seminar where they’re told, ‘This is what AI is.’ I mean bring some of what you’re doing right now into these workshops and say, ‘Let’s work on this,’” Spector advised.
Bakhshian said the most important lesson for energy professionals to understand about AI “is that AI is not simply a tool for automating tasks. It is becoming a decision-support technology. The energy industry generates enormous amounts of data, but data alone has limited value unless it can be transformed into actionable insights.
“Success will not be determined by who has the largest AI model or the most data. It will be determined by who can build AI systems that are trustworthy, explainable, physics-consistent and scalable,” she said.