The Evolution of AI in Industry
As our industry’s knowledge grows, so does the need for digital tools that help us navigate the complexity of our business, leverage the data we collect, and optimize the assets we operate. Among the most discussed technologies in recent years, one buzzword has been front and center following a breakthrough paper by 8 Google researchers in 2017[1] - Artificial Intelligence.
This survey aims to capture your perspectives on AI and automation in the heavy oil sector. Your input will help us better understand where the industry stands on AI adoption and where the most value is being or can be realized. First, let's set the stage with some background information.
This paper presents the result of a study that was conducted through August 2024 to analyze the 40 existing waterflood projects to address these questions.
What is AI?
The term AI is constantly evolving. Once, rule-based algorithms used in applications such as on/off controllers were considered AI. Today, few would call this AI because it is widely understood and used. For this survey, AI refers to any data-driven technique that enables machines to mimic human decision-making, learning, and problem solving.
Why do we care?
Our world is driven by data. The volume and complexity of that data have outpaced our ability to process it manually. Think of real-time sensors in drilling operations, the increased number of sensors in processing facilities, robotics and drones for inspection and surveillance, and fiber optic sensing for reservoir characterization. Data is our key differentiator in a race to become more optimized and cost competitive. But this data challenge has been around for decades, so what's with all the buzz now? To answer that question, we need to look at the evolution of AI techniques.
Under the hood - techniques used in AI and potential applications
Rule-based algorithms (1950s)
Also known as expert systems, these use if-then rules for decision making. Despite being one of the simplest forms of AI, it is extremely useful for automating expert knowledge and is widely applied in the industry. Traditional applications include process control and diagnostics. Another use case is automating rules to interpret geological and petrophysical data for exploration.
Physics-based and numerical algorithms (1990s)
Reservoir, process, and financial simulations use mathematical functions to describe a component's behaviour. While they are very useful, in the real world, components rarely exist in isolation. To be truly useful, an AI must optimize given a set of inputs, functions, and constraints. AI-driven solutions often use simulation to generate synthetic data, which are used in conjunction with production data to solve optimization problems. Examples of system-level optimization include adjusting operating parameters to maximize field production, adjusting steam allocation to minimize SOR, and selecting reservoir management strategies to maximize ultimate recovery.
Machine learning algorithms (2010)
A common challenge in any optimization problem is the availability of quality data. Often, a variable cannot be described by physics-based models alone. Installing more sensors is an obvious solution to the lack of data; however, it is often impractical. In such scenarios, machine learning (ML) can be used which, unlike rule-based algorithms that learn from human knowledge, learn patterns from data. A simple example is a regression model, in which inputs are correlated with outputs. Due to the rapidly increasing amount of data we collect, ML models have become integrated in many parts of heavy oil operations. For example, soft sensors use real-time data to infer a variable where a physical sensor is impractical or too costly to install.
Deep learning (2015)
A subset of machine learning that uses neural networks and large training datasets to handle complex scenarios. Natural Language Processing (NLP), machine vision, and speech recognition are all products of deep learning techniques that allowed AI to analyze modes of data that was not previously possible.
Generative AI (2018)
The paper published in 2017 revolutionized deep learning through a technique called the Transformer architecture and led to the current AI boom. The Transformer formed the basis for foundation models – large AI models that are highly adaptable without re-training. An important implication, Generative AI that can generate data such as text and images, meant that AI now has more advanced reasoning capabilities than ever before. [2] Another implication is the rise of general-purpose AI thanks to Large Language Models (LLMs) like GPT-3. The third implication is that essentially anything can be turned into useful data with AI, from unstructured text containing key metrics to core samples with valuable geological information. Multimodal foundation models like Google’s Gemini allowed images, text, video, and audio to be processed together, making AI highly versatile and accessible[3].
Agentic AI (2023)
Combining all the techniques before, the concept of an AI agent with human-like autonomy becomes a reality. The power of AI agents is that they are no longer restricted to specific tasks, but can adapt, orchestrate, plan, and execute a multitude of tasks.
Is the hype real or smoke?
With AI developing at an unprecedented pace, it’s fair to question whether emerging AI solutions are actually transformative or just overhyped. The Hype Cycle concept developed by Gartner offers a helpful lens. It describes how technologies tend to go through a few key phases, including a peak of inflated expectations, followed by a trough of disillusionment, and settling into a plateau of productivity[4]. While many AI technologies may still be in the peak inflated expectations, it’s important to realize that there is potential for sustained benefits once we understand their capabilities and limitations.
Tell us what you think
To help us gauge how the AI evolution is playing out in the real world, we invite you to answer a few short questions.
Click here to take a short survey (3 min)
References
[1] Vaswani, Ashish, et al. "Attention is all you need." Advances in neural information processing systems 30 (2017). arXiv:1706.03762.
[2] Veronika Samborska (2025) - “Scaling up: how increasing inputs has made artificial intelligence more capable” Published online at OurWorldinData.org. Retrieved from: 'https://ourworldindata.org/scaling-up-ai' [Online Resource]
[3] Li, Chunyuan, et al. "Multimodal foundation models: From specialists to general-purpose assistants." Foundations and Trends® in Computer Graphics and Vision 16.1-2 (2024): 1-214. arXiv:2309.10020.
[4] Gartner. "Interpreting Technology Hype." Gartner, 2025. Accessed April 9, 2025. https://www.gartner.com.
AUTHOR BIO
Yunji (Jimmy) Jiang, Innovation Engineer, Cenovus Energy
Jimmy Jiang, serving as a Director at the Canadian Heavy Oil Association (CHOA), has a robust background in chemical engineering marked by a commitment to innovation and digital transformation within the energy sector. Jimmy is currently an Innovation Engineer at Cenovus Energy, where he drives forward-thinking initiatives that align with the evolving demands of the energy landscape. Having a strategic mindset, Jimmy understands the critical role of digital innovation in enhancing operational decision making, safety and sustainability in heavy oil production.
Jimmy’s work with the CHOA underscores his dedication to fostering a collaborative environment for knowledge sharing among professionals. He has been involved with the CHOA Editorial Committee and volunteering as student coordinator at the CONNECTS conference. In 2025, Jimmy plans to continue advocating for technology and innovation in the heavy oil industry by highlighting the positive impacts of technology adoption in the CHOA Journal, fostering partnerships at CHOA events, and supporting emerging talent initiatives.
Outside of work, Jimmy is a close-up magician and loves to engage with the local art community, regularly participating in cultural events and volunteering at the Calgary International Film Festival. He also enjoys backpacking through both mountains and cities around the world.