Key takeaways
- AI adoption in the energy industry has shifted from experimentation to competition. Most companies now accept that AI is essential; the key question is no longer whether to use it, but how to deploy it better than competitors.
- Two strategic camps are emerging: “incrementalists” and “revolutionaries.” Incrementalists are using AI to improve existing workflows (faster analysis, anomaly detection, automation), while revolutionaries are aiming to redesign the entire operating model around AI-native capabilities.
- AI has the potential to fundamentally reshape how energy companies operate. AI-native approaches can democratize expertise, integrate planning and execution loops, and enable continuous probabilistic decision making for capital allocation and risk management.
- Waiting to adopt AI may create a widening competitive gap. Because AI systems improve with data and use, early adopters gain compounding advantages, meaning companies that delay implementation may struggle to catch up.
As the hype around artificial intelligence (AI) intensifies, it has been important to test our understanding of what's happening at the coalface—to separate the signal from the noise and, frankly, to make sure we're not falling prey to our own favorite theories.
I've spent much of the past year on the road, talking with customers. Most of the senior leaders I speak with understand that AI is here to stay. The conversation has moved beyond "Should we do this?" to "How do we do this better than our peers?" That shift is significant.
Two years ago, I sat in boardrooms where AI was treated as a curiosity or, worse, a distraction from the real business of extracting hydrocarbons. Fast forward to today and it’s hard to have a conversation where AI isn’t the central topic. Sometimes, it’s the only topic.
The debate is no longer about relevance; it's about ambition. But beyond that basic alignment, two different camps are emerging, each with a very different view of what AI might mean for their companies. Understanding which camp to be in is one of the most consequential questions facing energy leaders today.
The AI incrementalists
The first group is the AI “incrementalists”. These companies are implementing AI in their existing workflows, largely recognizing that the technology is more powerful than any other introduced in the digital era. The efficiency gains they’re seeing are real and measurable.
Reservoir characterization tasks that once consumed weeks of specialist time are being completed in hours. Field development planning teams are exploring far more scenarios than previously feasible, compressing the economic uncertainty in execution-phase decisions.
Similarly, production and drilling anomaly detection—which historically depended on an experienced engineer being in the right place at the right time—is now continuous and taking place in real time.
The improvements, in many instances, are no longer incremental. Yet still, AI is essentially a bolt-on to the existing operating model. The organizational structure hasn't changed. The planning cadence hasn't changed. The relationship between departments hasn't changed. AI is making the machine run faster, but the machine itself is the same one we've been operating for decades.
For many companies, this is a perfectly rational starting point. But it's not the whole story.
The AI revolutionaries
There's another group of companies asking a fundamentally different question. Not "Where can I apply AI to my current processes?", but rather "What would our industry look like if we designed it from scratch with these capabilities in our hands from day one?"
These leaders see AI not merely as a technology to adopt but as a forcing function; a pressure that is driving their organizations to align with a version of themselves 10 years in the future. Greater returns, lower costs, and radically different organizational structures.
For some, this is uncomfortable. For others, it's liberating. As one CTO I spoke to said, "Every other technology change let us improve on what we already did. This one is forcing us to ask whether what we are doing is even the right thing."
When these leaders describe their ambition, a common thread emerges. They see AI as the means to continuously optimize not just parts of the system but the whole system as it learns. That's not an incremental curve—it’s a different kind of trajectory entirely.
Three dimensions of this reinvention stand out.
Inverting the talent model
Today, most operators staff for peak complexity and carry that expertise continuously, whether it's being fully utilized or not. The cost is enormous, while the model is increasingly being called into question.
The experienced workforce is shrinking. Retirement rates are outpacing recruitment. The average age of a petroleum engineer is climbing, and the industry is struggling to attract new talent from a generation that sees more glamour in technology companies than in oilfield services.
With an AI-native model, the equation inverts. Deep domain expertise gets encoded once—drawing not only from within the bounds of a single organization, but also from decades of accumulated industry knowledge—and deployed everywhere. A mid-tier operator working a frontier basin gains access to the same quality of decision support as a supermajor running a mature field. A new graduate connected to an army of skilled AI agents can be onboarded and made productive in a fraction of the time it takes today.
Once again, I must reiterate: This isn't an incremental improvement in workforce efficiency. It's a restructuring of who can compete and where. The democratization of expertise is changing the competitive landscape of the industry itself.
Collapsing the planning and execution cycle
The oil and gas industry has always lived with a painful separation between subsurface planning, drilling execution, and production optimization. These are typically run as sequential, departmental handoffs: one team builds the model, another team drills the well, a third team manages production. Information flows between the teams imperfectly and with significant latency.
Agentic AI—systems that don't just analyze but act, orchestrate, and learn—creates the connective tissue to run these as a continuous, integrated loop rather than a linear relay. The well being drilled informs the completion design in real time. The production data from last year's wells is used to identify where to drill next quarter. What was three departments throwing work over walls becomes one cohesive, self-improving system.
The technology is being used to handle tasks below the line of genuine intellectual value, so that human expertise can operate at the level it was always meant to. The engineer who spent Monday morning manually checking production data can instead spend that time thinking about what the data actually means across the portfolio. We’re effectively getting closer to the decades-old dream of the truly integrated digital oilfield.
Rewriting the risk and capital allocation model
Today, final investment decisions in exploration and production are still largely built on deterministic thinking with uncertainty bolted on. A team runs a handful of carefully chosen scenarios and commits capital on the basis of a single expected outcome framed by sensitivity analysis. It's a process designed for a world in which computing power was scarce and modeling was expensive.
AI-native workflows have eliminated this constraint. Today, operators can run thousands of probabilistic scenarios continuously. In turn, capital allocation has transformed from a periodic, high-stakes decision into a dynamic, continuously optimized process. Confidence intervals have changed as a result, as has the ability to detect emerging risks. Companies simply don’t view or deal with uncertainty the same way they have in the past.
It's not about making better spreadsheets. It's about changing the industry's fundamental relationship with risk. In a world where all data is always available and meaningful, the companies that cling to periodic, deterministic decision making will be outcompeted by those that have learned to operate in a continuous, probabilistic mode. The latter will simply make better bets, more often, and at a lower cost.
Is it possible to pick a “right” side?
Unfortunately for energy leaders, the decision of which camp to be in (the incrementalists or revolutionaries) isn’t so simple. They must be in both, with each side posing its own unique dilemma.
The achievements of the incrementalists will rapidly become table stakes or the minimum standard required to remain competitive. Any operator not actively deploying AI into reservoir characterization, anomaly detection, and workflow automation today is already falling behind. These aren’t optional investments; they’re the cost of doing business. The revolutionaries, meanwhile, will experience expensive failures. They will suffer organizational turmoil. Some of their bets won’t pay off in the timeframe they hoped.
Grasping the nettle, however, isn’t a choice. Because much more so than previous waves of technology (e.g., cloud, Internet of Things, digital twins), AI represents a genuine opportunity for operators to create durable, structural differentiation from their peers. And in an industry where competition for capital is becoming ever more intense, where investors demand discipline and returns above all else, that differentiation will increasingly determine who attracts investment and who doesn't.
Many of the leaders in the second camp see this clearly and believe there will be less room for average performers in the future oil and gas business. Only those companies eager to rise above their peers will secure their future.
One way that companies can intelligently move ahead in today’s environment is to employ a portfolio approach. This entails funding efficiency plays first to generate near-term returns and build organizational credibility, and then ring-fencing a separate effort for reinventing the organization’s operating model (an effort with separate metrics and a fundamentally different tolerance for failure).
The key is to build both strategies on the same underlying technology platform. Doing so will allow the table-stakes work to fund and inform the “moonshots” and help determine how new ideas can eventually scale. Without a common platform, organizations will end up with pilot projects that never advance and innovation labs that never connect to the business.
The overarching goal is to avoid letting the first group's governance kill the second group's ambition.
The “fast follower” trap
There is one more dimension to the AI adoption topic that’s relevant to any company still in "wait and see" mode.
In previous generations of technology change, it was almost always rational to be a “fast follower”. Let the early adopters pay the learning tax, fix the bugs, absorb the integration pain, and then swoop in when the costs come down, and the playbook is derisked and established. The energy industry, understandably cautious with capital and intolerant of operational risk, built this approach into its institutional DNA.
Broadly speaking, it has been successful. But the same strategy may not work this time around.
AI systems improve with use. They learn from data, from feedback loops, from the accumulated experience of every decision they help inform. Which means the operators that build AI-native workflows today don’t just get a head start; they enter a compounding cycle whereby systems improve and generate more revenue to fund further tech investment.
The models get smarter. The agents get more capable. The institutional knowledge base deepens. And critically, the humans working alongside these systems develop an intuition for how to use them—a competitive advantage in and of itself.
A small lead today may become a significant advantage in 12 months and a potentially insurmountable one in three years. Kurzweil's Law of Accelerating Returns is playing out in the operational data of companies that have committed early. The competitors who assume they can catch up later are making a bet that the pace of change will slow down. But there’s little evidence that it will. If anything, the emergence of agentic AI systems that can orchestrate complex, multistep workflows with minimal human intervention is compressing the timeline further.
The leadership mandate
What practical steps can leaders take to ensure that their organizations don’t fall behind?
The reality is that technology is the easy part. The hard part is change management.
Every strategy starts with conviction. The transition to AI-native operations requires leadership to redesign incentives, break down departmental walls (just as the integrated agentic loop demands), and tolerate the short-term turbulence that always accompanies genuine transformation. Half-measures simply won’t be enough.
Second, they must explicitly manage both camps as a portfolio. Incrementalist initiatives need clear return on investment targets and rigorous execution discipline. Revolutionary initiatives need protected funding, different success metrics, and permission to fail because some of them inevitably will. The two efforts require fundamentally different management approaches.
Third, companies must invest in a common platform beneath both camps. The domain-specific AI layer—the models trained not just on generic data but on decades of physics, engineering judgment, and operational context specific to the subsurface—is where competitive advantage compounds. This is not a capability that can be outsourced to a hyperscaler or built from scratch through a start-up. It requires deep domain expertise encoded into technology, and it's the layer that turns generic AI potential into specific, defensible, operational value.
Last but not least, leaders must change what they ask their best people to work on. Responsibilities of talented engineers and geoscientists should be shifted away from tasks that AI can already do toward the judgment-intensive, creative work that is better suited to humans. That cultural shift of redefining what human expertise is for takes longer than deploying the tools. So, it has to start first.
The reshapers and the reshaped
The energy industry stands at a genuine inflection point. Not the kind that gets declared at every conference and then forgotten six months later, but the kind where the decisions made in the next one to two years will shape competitive positions for a decade or more.
The incrementalists are doing important work, and their gains are real. But the revolutionaries are pointing toward something bigger, even if the path is uncertain and expensive. The companies that will define the next era of this industry are the ones with the strategic imagination to pursue both approaches simultaneously, with the discipline to deliver near-term returns, while maintaining the courage to bet on fundamental reinvention.
My opinion—based on experience working with customers across different geographies and operating environments—is that AI isn’t just another technology hype cycle that will quietly deflate. The proof points are too concrete, the efficiency gains too large, and the competitive dynamics too powerful for that to be the outcome.
The early movers are already pulling ahead. The compounding has already begun. And the window for fast followers is closing faster than most people in the industry want to believe.
