“Any sufficiently advanced technology is indistinguishable from magic,” wrote Arthur C. Clarke—cowriter of the screenplay for 2001: A Space Odyssey—in a 1968 letter to Science magazine. This after Isaac Asimov (another of the “Big Three” science fiction writers) had proclaimed in 1952 that “an uninformed public tends to confuse scholarship with magicians."
What these tech visionaries meant by these statements is that, in the real world, the meaningful and lasting impact of innovation is not conjured up out of thin air. It doesn’t just happen. On the contrary, it’s only ever the result of the painstaking work of brilliant and persistent people.
Useful technology must be engineered.
This is something I’ve seen our industry do successfully again and again. The great advances I’ve witnessed in my professional life have come through someone spotting a problem and engineering a powerful solution.
I will never forget the excitement in the operator’s eyes when we delivered virgin fluids from the reservoir to them for the first time, making their dreams a reality. The modular formation dynamics tester remains an astonishing and indispensable piece of tech to this day.
And how about directional drilling? A marvel of engineering that enables drilling multiple wells from a single location, hitting a one-square-meter target thousands of meters away, thousands of meters underground. Or we could talk about 4D seismic providing dynamic insights into reservoir behavior, thereby optimizing production and increasing recovery.
It’s all awe inspiring. And it proves that the future has always been engineered by exceptional minds in the energy sector. This isn't luck. And, although it can sometimes seem like it, it’s certainly not magic either.
When we entered the digital era, things that only a few years earlier were considered pure fantasy became possible. The technological leaps we now see leave us openmouthed with wonder. But when we look at the underlying patterns, we understand that success only comes with purpose.
Nowhere is this more relevant now than in the realm of artificial intelligence (AI).
We’ve heard about so many types of AI: narrow AI, general AI, superintelligent AI, robotic AI, conscious AI, and of course, generative AI. I want to propose a new label for a flavor of AI tuned for the demands of our industry. I call it “engineered AI.”
Engineered AI has a deliberate bias for our industry, specifically to solve real-world problems at a massive scale and for significant impact.
To deliver this critical domain context, models must be trained with energy-specific data and integral knowledge of our industry. This is not artificial “general” intelligence; it’s a highly distinct intelligence, unapologetically optimized for the energy sector.
And engineered AI means that the foundations are robust. More accurate. More secure. More scalable. This cannot happen without collaborations that support the investment required to create a fresh data infrastructure for the industry. We’ve never had a better reason to address the age-old conundrums around siloed and locked up data and connect data sources right across the value chain, from subsurface through field development to drilling and production. And to do all this in a way that ensures trust and transparency, with insights illuminated by AI.
To be clear, engineered AI is not just an idea—it's something that’s already happening today. And I can give you a great example.
Production operations consume a huge amount of cash, linking their activities directly to business performance. But the interconnectedness of physical production systems (the reservoir, wells, pipelines, and facilities) is often not mirrored in the digital systems that underpin them. As a result, holistic system optimization is often difficult—if not impossible.
At this very moment, there’s a group of oil fields comprising about 270 wells distributed across the jungles of South America. These mature fields are prone to performance issues related to flow assurance, lift system failures, and waterflooding. Their waterflood design is already quite sophisticated, using cloud computing and high-fidelity numerical simulation models on cloud-based software platforms.
In a truly dynamic production environment, however, there’s a need to continuously tune the waterflood design in response to field operating conditions to maximize reservoir performance. Using numerical models to do so would require too much effort, time, and resources—but engineered AI can solve that.
In fact, for these wells, it already has.
A set of AI-driven waterflood models that combine machine learning and physics-based capacitance resistance models were developed specifically for this purpose. This approach accelerates operational waterflood decisions by generating insights that allow for the type of rapid action expected of a dynamic operations environment.
But operations can be tricky in the Amazon. These are some of the world’s most remote onshore well locations. Hence, executing on the prescribed waterflood is often not straightforward at all. Here’s where engineered AI strikes again. Through a combination of cloud and edge platforms—infused with AI, of course—fully autonomous well pads have now been created.
Humans barely need to visit. The positive economic and environmental impacts are obvious.
What does this mean? It means that actions such as performance optimization of electrical submersible pumps (ESPs), adjusting waterflood pumps, and chemical injection management are orchestrated with little to no human intervention. There’s an autonomous interplay between the cloud and the edge to ensure a balancing act that delivers peak performance 24 hours a day, 7 days a week. Not to mention that physical and virtual flowmetering are also continuously monitoring for production underperformance, while reliable simulation workflows identify corrective actions for managing shortfalls.
These corrective actions (think actions such as adjusting well chokes or ESP parameters) are remotely and autonomously executed, triggering the waterflood tuning workflows and, in turn, generating injection recommendations that are again implemented remotely and autonomously. To ensure that flow assurance issues are kept at bay, the chemical injection system autonomously adjusts in response to these injection and production changes.
And it doesn’t end there. To deliver more than just incremental value from the entire system, we need to reach with AI into pipeline networks and facilities, too. If we don’t have those optimized, it won’t matter how much the well pad produces through the optimization of its reservoir and well operations because we still won't be able to deliver the maximum number of barrels at the end of the value chain.
The good news is that baking engineered, physics-informed AI into the platforms can enable pipeline and facilities debottlenecking, so that operators receive the full throughput of their wells, pads, and reservoirs into their facilities.
AI can do that if it’s engineered correctly. An AI model developed using training data from robust physics-based simulations, for example, rapidly generates flow assurance risk profiles. The associated corrective actions for the entire network (even with more than 800 km of pipeline) are taken within a matter of minutes, all of which can be implemented through the edge infrastructure.
In other words, modern tech has reached whole system optimization—from pore space to market.
What’s the potential impact of all this? Results such as
- sustained 4% increase in production
- 60% increase in people efficiency
- 25% reduction in well and equipment failure index
- near-perfect chemical injection system reliability
- and, most significantly, the potential to see a 57% reduction in associated CO2 emissions.
This is what I mean by “engineered” AI. Solid foundations that can scale. AI tuned for the oil field. Using new digital tech to move the dial. Because when we get it right, AI isn’t an economic threat, it’s a massive opportunity. New jobs. New value creation. New energy trilemma solutions.
