In today’s industry, many say geophysics is not needed because:

  1. There are too many wells, so it is easy to map the subsurface.
  2. Seismic does not have the resolution to see the thin sands.
  3. Have not had a lot of success with geophysics in this play.

This sentiment is further expressed when artificial intelligence (AI), predictive or prescriptive analytics, or machine learning (ML) are discussed, with many feeling geophysicists (geoscientists) will soon be replaced with AI or ML.

Despite what is being said, seismic will still be needed to understand what is occurring between the wells, and more emphasis will be placed on it with the implementation of AI, predictive or prescriptive analytics, or ML. As regards the elimination of jobs, many predict AI, predictive or prescriptive analytics, or ML will just be applied to routine tasks, so individuals can undertake more intricate and imaginative positions that necessitate greater levels of proficiency and knowledge. Domain expertise will definitely be required in order to select the right attributes or features for the ML or predictive or prescriptive analytics (Schulte et al., 2023).

When geomodelling first began, it was realized that it was difficult to use the seismic with well log data or rock properties because (Schulte, 2022a):

  1. Difference of sampling between wells and seismic
  2. Difficulty in properly modelling the petro-elastic relationships
  3. Deterministic inversion produces average impedances over intervals
  4. Tends to exaggerate reservoir connectivity
  5. Underestimates net reservoir volumes.

To fix some of these issues, rock properties volumes created by Neural Networks (NN) were utilized. The NN can estimate any elastic constant or well log, such as Vp, Vs, density, Vp/Vs, porosity, saturation, etc., using multiple seismic attributes with various weights. By using multiple samples from a multitude of attributes with various weights, NN removes the effect of the wavelet, like what happens in inversion; and the NN can be computed at sub-seismic rates, which increases the resolution beyond what is achieved with seismic inversion (Schulte et al., 2023; Ross, 2016). The NN rock property volumes also have fewer smearing effects from the seismic wavelet (Blouin and Gloaguen, 2017; Lancaster and Whitcombe, 2000), and there is no leakage of the model into the volume.

There are two ways to include seismic such as the seismic facies volume into the kriging of the data in the geomodel by using (Schulte et al., 2023):

  1. Co-kriging, which takes advantage of the covariance between two or more regionalized variables that are related when the main attribute of interest (well data) is sparse, and the related secondary information (seismic) is abundant.
  2. Kriging with external drift, which has an assumption that the primary variable is linearly related to the secondary variable.

Still, there has not been the move towards digitalization and the acceptance of AI, predictive or prescriptive analytics, or ML many thought there would be. To understand the resistance to AI, predictive or prescriptive analytics, or ML in the oil and gas industry, we need to examine the Beckhard and Harris (1987) change equation which is:

Dissatisfaction x Desirability or vision x Practicality > Resistance to Change

Equation 1: Beckhard and Harris (1987) change equation.

Deloitte’s Digital Maturity Index shows oil and natural gas industry is lagging other large industry verticals (Digit, 2021), due to the resistance which could be:

  1. Oil and natural gas industry being traditionally slow-moving on projects due to profits being earned over years and decades, and oil and natural gas operators do not want to deviate that much from their core operating model (Digit, 2021)
  2. Some asking why change; it has always been done this way especially when Energy (oil and gas) was the best performing sector on the S&P 500 in 2022 due to the high price for oil and natural gas (Toppe, 2022).

The industry needs to acknowledge that the digital transformation that many articles discuss and that boards of directors have been talking about implementing is not an easy task to do. It will take a Chief Transformation officer being named to the board, who is an extension of the CEO.

One of the good things that has happened in oil and gas with the high prices of oil and gas and the practise of capital discipline is that many oil and gas operators have excess cash on their balance sheets, which could cause them problems with investors due to investors beginning to ask about opportunity costs, which are the benefits lost by choosing one option over another.

This may lead to investors asking oil and gas operators to invest this excess cash into digital transformation, which could (Digit, 2021; Kolaczkowski, et al., 2021):

  1. Save the oil and natural gas industry over $70 billion a year in operating costs,
  2. Lower emissions
  3. Increase productivity.

AI, predictive or prescriptive analytics, or ML could also attract students into the industry which many do not consider. Attraction of students to the industry is highly desirable due to 147,000 geoscientists being expected to retire in the U.S. by 2026, and only 62,000 geoscience students graduating with bachelor’s, master’s and doctoral degrees to fill those gaps (Saucier, 2020). Many students may be enthused to work on these types of projects involving AI, predictive or prescriptive analytics, or ML since they can be challenging and their participation in such projects looks good on a resume especially if some of the results leads to reducing costs and reduction of emission of greenhouse gases (GHG) while producing oil and gas. They can quote how much was achieved by implementing such a project on their resumes.

It can also alleviate some of the stress on the asset teams due to the inability to attract talent, which will cause shrinkage within oil and gas companies. This shrinkage of available personnel may cause oil and gas companies to be limited in pursuing new projects outside their core areas.

To also help attract students into oil and gas, oil and gas operators need to become one of the “Fortune’s 100 Best Companies to Work For.” These companies tend to be the best performing organizations on the S&P 500 (Goenner, 2008), and they also attract the best talent from universities.


Paraphrasing Samuel Clemens, “The reports of the death of geophysics is greatly exaggerated.”

Many times, it is heard that geophysics is not required in this play because there are so many wells, the seismic does not have the resolution, or there hasn’t been a lot of success with geophysics in this play, etc. This sentiment is further expressed as the industry talks about Artificial Intelligence (AI), predictive or prescriptive analytics, or machine learning (ML), with many feeling geophysicists (geoscientists) will soon be replaced with AI or ML. This notion is generating resistance against geophysics (Gartner’s Hype Cycle Methodology – CIO Wiki, n.d.; Gartner, Inc., 2021; Wikipedia contributors, 2023).

With the implementation of AI, predictive or prescriptive analytics, or ML, there may be a bigger push to use seismic and seismic attributes as features to understand what is going on between well locations and to help with more accurate predictions (Schulte et al., 2023). To choose the right attributes for AI, predictive or prescriptive analytics, or ML to be successful, one needs to understand which features or attributes are correlated. We want to avoid using attributes that are correlated because correlated data creates redundant information, which skews the results in the regression model. What happens is that a change in one variable causes a change in another, causing the model to fluctuate a lot, and results to become unstable, varying a lot given a minor change in the data (Schulte et al., 2023; Wu, 2020). Deciding which attributes to keep and which to remove requires domain expertise (Schulte et al., 2023).

What may occur is that, with the implementation of AI, predictive or prescriptive analytics, or ML, the type of work that individuals conduct in an industrial sector will change, with the AI assuming more of the routine responsibilities, and individuals undertaking more intricate and imaginative analyses that necessitate greater levels of proficiency and knowledge. Thus, they see AI will be aiding workers and not replacing them (Ghadban, 2023; Lee, 2017; Liang et al., 2019).

To implement AI, predictive or prescriptive analytics, or ML, we need to understand where it falls on the Gartner Hype cycle for emerging technologies. The Gartner Hype Cycle for emerging technologies (Figure 1) supplies a high-level view of emerging technologies that should be tracked and monitored for their potential to solve real business problems and exploit new opportunities.

Figure 1: Hype Cycle for emerging technologies. This curve is broken up into the Technology Trigger, where a potential technology breakthrough triggers significant publicity, followed by the Peak of Inflated Expectations, where there are early success stories followed by scores of failures. This is followed by the Trough of Disillusionment, where interest wanes, and implementations fail to deliver. After that is the Slope of Enlightenment, in which the technology and its benefits are better understood. Finally comes the Plateau of Productivity, where mainstream adoption starts to take off (Wikipedia contributors, 2023). Taken from Dion (2020) and Lasopamenu (n.d.).

The Hype Curve shows how a technology will evolve over time, providing insight to manage the deployment within the context of an organization’s business goals (Gartner, Inc., 2021; Gartner’s Hype Cycle Methodology – CIO Wiki, n.d.; Wikipedia contributors, 2023).

The Hype Cycle is divided into five distinct phases (Wikipedia contributors, 2023):

  1. Technology Trigger – potential technology breakthrough triggers significant publicity.
  2. Peak of Inflated Expectations – early success stories followed by scores of failures.
  3. Trough of Disillusionment – interest wanes, and implementations fail to deliver.
  4. Slope of Enlightenment – technology and its benefits are better understood.
  5. Plateau of Productivity – mainstream adoption starts to take off.

In this Hype Curve, predictive analytics is entering the Plateau of Possibility phase, where it is becoming mainstream. Prescriptive analytics is approaching the Trough of Disillusionment, where interest wanes and implementations fail.

The goal of this paper is to examine the implementation of digitalization and AI, predictive and prescriptive analytics, and ML in oil and gas and how it needs to be implemented successfully. This leads to leadership, change management and the adaptation of transformational leadership. Part of it is technical, and part is applying management skills. It comes back to this looking simple, but it is much more complex.

Implementing AI, predictive and prescriptive analytics, and ML is a large undertaking within any organization. It takes a large effort, and it can be daunting at times, but the results are worth it.

The Beckhard and Harris (1987) change equation is:

Dissatisfaction x Desirability or vision x Practicality > Resistance to Change

Equation 1: Beckhard and Harris (1987) change equation.

Dissatisfaction – the amount of carbon being emitted; disappointment in the production; overrun on costs compared to the competition.

Desirability or vision – predictive analytics can be done in reduced time compared to a reservoir simulation. Time is money, and the organization can make a crucial economic decision before the competition. The vision is that digitalization could save the oil and natural gas industry over $70 billion a year in operating costs, could lower emissions, and could increase productivity (Digit, 2021; Kolaczkowski et al., 2021).

Practicality is:

A. Software is available through third parties to make the transition easier.

B. Proprietary expertise can be brought in through hiring and used to build internally the necessary tools for AI, predictive or prescriptive analytics, or ML.

C. Outsourcing to companies that specialize in AI, predictive or prescriptive analytics, or ML, taking advantage of the years of industry expertise available in such companies.

D. Investment into startups and R&D.

Resistance can be multiple factors:

A. Employees may fear loss of jobs, affecting morale across the company.

B. The company is making profits currently without digitization, so why implement it.

C. Large upfront investment to develop.

D. Requires implementation of change management.

The change equation is a nice and straightforward way to demonstrate how to overcome resistance towards AI, predictive or prescriptive analytics, and ML.

If the industry reflects on it, the resistance must still be stronger than dissatisfaction x desirability or vision x practicality because, according to Deloitte’s Digital Maturity Index, the oil and natural gas industry is lagging most other large industry verticals (Digit, 2021). There is are ways to go on this digitalization, and many in this industry will be faced with it in the coming years because it is the direction in which our industry is headed. To facilitate digitalization, we need to implement change management. The goal of change management is to transition smoothly and efficiently, and it is best done if it is planned to avoid disruptions and employee resistance (Coursera, 2023). The positive motivation for going towards digitization is that it (Correa, n.d.):

  1. Creates cultural disruption towards innovation. It changes the way people work.
  2. Opens a door for new business opportunities, enabling the creation and development of new services and products.
  3. Improves operational efficiency and productivity.
  4. Generates a competitive advantage for the company.

Why most boards of directors want to do this digitalization is fear of disruption by tech-enabled competitors; the digital transformation should be led at the board level rather than given to the IT department (Correa, n.d.; David and Farzan, 2021).

Why leadership is important

Using Porter’s five-forces of industry, there are barriers to entry into the oil and natural gas industry, which are extremely strong due to high startup costs, high resource ownership, patents, etc., which have caused the industry to be sheltered from competition and new market entrants. This has prevented change and disruption from being forced onto the industry in the same manner as other industries, such as finance (The Investopedia Team, 2022; Digit, 2021).

Even when the price of oil has been high, oil and natural gas have been slow to modernize and invest in wholescale digital transformation. The major issue is that the oil and natural gas industry has traditionally been slow-moving on projects due to oil and natural gas earning their profits over years and decades, so the industry does not want to deviate that much from its core operating model (Digit, 2021). For those reasons, it has a risk-averse culture and slow and poor innovation-management practices.

Leaders in oil and gas are currently being challenged to transform themselves and their organizations to succeed in a changing world, including the push for low-carbon projects (Derkach et al., 2023). The goal is to look at how this can happen, especially through embracing new technology to solve some of these issues. Leadership needs to figure out how best to implement these necessary changes successfully.

The reason to learn more about predictive analytics and why it should be implemented is that the goal of predictive analytics is to reduce the risk of making investments that turn out to be unprofitable. It also allows us to see how small changes can add up to big impacts, given the fact that oil and natural gas projects operate over years (Schultz, 2021; Cann, 2020).

“It has always been done this way.”

Part of implementing innovative ideas and new ways of doing things is dealing with the concept of “why change; it has always been done this way.” To incorporate innovation, we need to be open-minded about how it can be used. There needs to be psychological safety where it is okay to say things openly, which fosters brainstorming. Just one individual will never develop ideas, it is usually done through brainstorming with a group, and of course, to learn new things, there must be leeway to make mistakes.

Another roadblock to digitalization is that data is siloed and fragmentated: the geoscience personnel may report to a Vice President of Geoscience and the engineering personnel report to a Vice President of Engineering (Kolaczkowski et al., 2021). Our goal should be to create an environment where it is safe for people to share data and ideas openly between disciplines and to take risks. This is what a learning organization is all about (Vector Solutions, 2022). It will require a multi-disciplined team to do things like predictive analysis, and it will encourage cross-disciplines to work together, breaking down siloes, brainstorming, and learning from each other (Schulte et al., 2023).

Moving towards digitalization and predictive or prescriptive analytics is not about replacing domain expertise. In fact, with digitalization, we need to rely more on domain expertise to be able to identify which features or attributes need to be selected for predictive or prescriptive analytics (Schulte et al., 2023).

The pace of the digital transformation is terribly slow in oil and gas, and that is causing students to avoid employment within the oil and gas industry, complicating the lack of enrollment in the engineering and geoscience programs that feed into the oil and gas industry. This is due to some students wanting to get involved in the challenges of AI, predictive or prescriptive analytics or ML because they see it as the future and want to work on projects involved with it so they can put it on their resume (Schulte et al., 2023; Venables, 2018).

AI, ML, Descriptive, Predictive and Prescriptive Analytics

These terms are used interchangeably all the time, so what are they?

AI is a wide branch of computer science that processes and analyzes data so it can understand and learn from past data points through specifically designed AI algorithms. ML is a subset of AI that uses statistical techniques and data to extract algorithms and models for learning. Descriptive Analytics gives an account of what has happened in the business with historical data, predictive analytics uses historical data to make predictions about the future, and prescriptive analytics uses historical data to advise us on what to do (Farmer, 2021; Tableau, n.d.).

Implementation of AI, predictive and prescriptive analytics, and ML

ML is being used especially with Neural Networks (NN) to build missing well curves resulting from cost limitations or borehole problems. Many have used Hampson-Russell’s Emerge to do this in their day-to-day operations.

We are seeing AI and NN being used in seismic processing to identify acquisition noise for noise attenuation. We can utilize modelled noise from 3D full wave modelling to train the neural networks or AI (Figure 2). This requires 3D seismic forward full wave modelling, which is generally used to parameterize seismic acquisition and processing.

Modelling is important in our work because it allows us to understand the consequences of our proposed decisions before they are implemented. Sometimes if the consequences are not understood, it can be costly.

Figure 2: 3D full wave modelling can be used for multiple reasons. The goal is to understand the consequences of our proposed decisions before they are implemented. The modelling can look at the size of the vibrators used, especially in environmentally sensitive areas. It can look at the vibrator parameters for slip-sweep acquisition, compare Prestack Time Migration (PSTM) versus Prestack Depth (PSDM), etc. All of this comes back to reducing costs yet maintaining the signal that is required.

Fault and horizon picking

With 147,000 geoscientists expected to retire in the U.S. by 2026, and only 62,000 geoscience students graduating with bachelor’s, master’s and doctoral degrees to fill those gaps (Saucier, 2020), oil and gas companies need to create and implement tools that will alleviate stress on asset teams that will have fewer individuals. It will decrease turnaround time to pick well locations. Seismic Interpretation software companies are building AI programs to pick faults and seismic horizons. This will speed things up.

Some are using fault probability or likelihood to build fracture and fault networks. To do this, we need to utilize edge-preserving structure-oriented filtering for preconditioning to illuminate the discontinuities. We have developed a workflow using the spectrally enhanced Intercept or Near Offset Stack, coloured inversion, and edge-preserving filtering, which produces acoustic impedance with higher frequencies (high resolution) and crisp fault boundaries (Figure 3) (Schulte et al., 2019).

Using AI to pick faults can pick up on discontinuities and build an accurate discrete fracture network (DFN) for use in the geomodelling or reservoir simulation. Some AI programs can be trained and then retrained, and as they learn, they may notice faults not observed by the geoscientist.

Figure 3: Suggested workflow to highlight discontinuities in the seismic data.

With AI, there is repeatability and consistency, despite personnel leaving or being rotated to other assets. This allows us to maintain a consistent product for predictive analytics or proxy modelling and reservoir simulation. Repeatability and consistency are required for regulatory boards and audits to determine the size of reserves in place.

NN to create rock property volumes

To incorporate seismic with well data or rock properties we need to utilize rock property volumes created by a neural network. The NN uses a network of functions to translate the seismic data and attributes into a well log or rock property volume (DeepAI, 2020). Using multiple samples from a multitude of attributes with various weights, removes the effect of the wavelet, similar to inversion without the issues of:

  1. Smearing effects of the seismic wavelet (Blouin and Gloaguen, 2017; Lancaster and Whitcombe, 2000)
  2. Leakage of the model

The NN is computed at sub-seismic rates, which increases the resolution of the NN volume beyond what is achieved with the seismic inversion (Schulte et al., 2023; Ross, 2016).

With NNs, we need to look for attributes that are correlated, which leads to the creation of redundant information (Figure 4). Redundant data skews the results in the regression model, causing a change in one variable to create a change in another creating the model to fluctuate a lot, and the results become unstable, varying a lot given a minor change in the data (Wu, 2020; Schulte et al., 2023).

Figure 4: Workflow of the neural network taken from Schulte (2022) and (Schulte et al. 2023).

With the geomodel there are two ways to include the NN rock properties, such as the seismic facies volume, into the kriging of the data (Schulte et al., 2023):

  1. Co-kriging, which takes advantage of the covariance between two or more regionalized variables that are related when the main attribute of interest (well data) is sparse, and the related secondary information (seismic) is abundant.
  2. Kriging with external drift, which has an assumption that the primary variable is linearly related to the secondary variable.

How to make AI, predictive and prescriptive analytics, or ML successful

When trying to implement AI, predictive or prescriptive analysis or ML, the biggest hurdles are (McKinsey & Company, 2016):

  1. Obtaining an effective analytics program
  2. Lack of leadership support and communication
  3. Ill-fitting organizational structures
  4. Troubles finding the right people for the job.

When looking at companies that are successful and effective with data and analytics, it comes back to senior management being involved with data and data analytics. With the low performers of analytics, they say the biggest hurdle is designing the right organization structure to support analytics (McKinsey & Company, 2016).

In some of the oil and gas companies there is already a board position called Chief Transformation Officer or CTO. These individuals sit on their executive leadership team and inspire and function as role models for the sort of behaviour needed for change. Their compensation should be linked to performance, and they need to function as an extension of the CEO. In Repsol, the CTO is the Deputy CEO. CTOs have a strong cross-functional background and have been in a variety of different business situations and challenges during their career (Gorter et al., n.d.).

This transformation will also require change management, which requires a clear framework for implementing the process, providing a roadmap for change that outlines the steps, processes, and activities required to achieve the desired outcomes. Using a framework facilitates communication between stakeholders and creates a shared understanding of the change initiative. It helps to reduce confusion, minimize resistance, and increase collaboration between teams (Abbas, 2023).

Developing brownfields rather than drilling exploration green fields

New oil or gas fields take an average of 5.5 years from discovery to first production and another 17 years to reach peak output (Wachtmeister and Höök, 2020) and with the implementation of government policies it will cause half of the new cars sold in U.S. to be electric and 60% in Canada (Boudway, 2022; The Canadian Press, 2022). This has created uncertainty about the demand for oil in the future.

Many investors do not want oil and gas operators to invest in new oil and gas fields; rather, they want them to focus on developing their current plays, re-investing only 50% of their profit into new drilling and using the other 50% of the profits to pay down debt, provide a reliable dividend to shareholders, and buy back shares (Worland, 2022).

New exploration for oil and gas has fallen sharply worldwide this year, with a total acreage of new oil and gas leases falling to near all-time lows, and in 2022 only 44 global lease sales were completed, the lowest level since 2000 (Nilsen, 2022).

Effect of capital discipline

Due to capital discipline, oil and gas operators are currently holding substantial amounts of cash on their balance sheets. There is currently a dispute amongst operators about how to use this excess cash. It can cause investors to ask why the money is not being put to use since there is an opportunity cost, which is the amount of the difference between interest earned on holding cash and the price paid for having the cash as measured by the company’s cost of capital (McClure, 2021; Valle and Kumar, 2023).

Oil and gas operators can investigate how this excess cash could be invested into digitalization and how it could affect our different plays. Machine learning approach was introduced to the petroleum engineering area over ten years ago to build faster production models (Mohaghegh, 2005, Mohaghegh et al., 2009; Kong et al., 2021) so its application in oil and gas is not new.

Shale plays (tier 2 and 3 rocks)

These are also known as basin-centered gas systems (BCGS) and have similar geology to the Deep Basin (Figure 5) where there is a series of stack pays of tight sands and shales with coals and shales acting as the source rock. The three most active and higher-producing basins for oil are the Permian, Eagle Ford, and Bakken. In the Permian and Eagle Ford, there has been a large drawdown of the drilled uncompleted wells.

Some believe the past success of the shale plays has been due to high grading, where oil and gas operators have been drilling their most productive prospects or sweet spots first. They feel that future increases in shale drilling productivity will be more a function of continued high grading and less of a function of ever-changing drilling and completion techniques. This will eventually cause the available drilling sweet spots to run out (Goehring & Rozencwag, 2019; Slav, 2022).

Most look at the first 2 to 3 years or the initial flush of a shale well production since during that time the well declines significantly but the amount of production allows for the well to be paid off which gives these wells a high return on investment (ROI) making them attractive to investors. Now we are realizing the importance of tail production where 75% or more of the well production will be achieved in the first ten years. This is causing many envisioning enhancing the tail production could be a potentially significant research area (Middleton et al., 2017). Predictive analysis could help with tail production to understand what could be done to enhance production such as re-fracking or utilizing water alternating gas such as supercritical CO2 which will improve the sweep efficiency (Schulte, 2022b)

Predictive analysis within the shale plays increases the speed of the entire analysis and modelling process, reducing 6 to 8 weeks of manual research down to seconds to find (Fu, 2018):

  1. Stimulation volume
  2. Fluid types
  3. Flow rate
  4. Well lateral and cluster spacing
  5. Other numerous considerations.

Application of AI, predictive or prescriptive analytics and ML in the shale basins have been done in the following projects:

  1. Deep neural network to forecast the cumulative production from Bakken shale (Wang et al., 2019; Kong et al., 2021)
  2. Principal component analysis, unsupervised clustering, and regression analysis were used to identify shale gas production patterns in the Marcellus formation (Zhou et al., 2014; Kong et al., 2021)
  3. Random forest algorithm and deep neural network was used to optimize the completion strategy in the Bakken shale formation (Luo et al., 2018; Kong et al., 2021)


With the Trans Mountain pipeline operational by 2024 and the capacity increasing to 890,000 barrels per day (bpd) from 300,000 bpd (Williams, 2023), it will create pressures to produce more heavy oil. Companies like Imperial are considering restarting expansion projects (Orland and Morgan, 2023).

With the oilsands, they have high up-front costs in the billions of dollars, and it takes years to build the mines or thermal projects that are required to extract oil sands bitumen, but the depletion rate is slow compared to shale plays, and it is a long-lived resource base. Currently, the oilsands operators are pumping as much as they can from existing facilities (Seeking Alpha, 2014; Nickel, 2021).

To lower break-even costs in the oilsands, we have been seeing smaller, incremental brownfield expansions rather than large-scale greenfield projects. The oilsands operators have not only indicated smaller expansions but plans to achieve cost savings by tying these into existing central processing facilities (CPFs) rather than building new CPFs (Rystad Energy, 2021).

Oilsands operators have been consistently lowering their break-even price, and it has fallen from $77.52 USD in 2015 to $45.92 USD in 2023 or 41% due to using this strategy (Kaplan, 2023).

AI and ML are being used in SAGD to:

  1. Forecast recovery in SAGD operations (Ansari, et al.,2020; Canbolat and Arturi, 2022)
  2. Optimize steam injection (Guevara et al., 2018; Canbolat and Arturi, 2022)
  3. Optimize steam allocation in a multi-pad SAGD reservoir model (Guevara and Trivedi, 2022; Canbolat and Arturi, 2022)
  4. Shale-barrier characterization (Kumar et al., 2020; Canbolat and Arturi, 2022)
  5. Forecast dynamic changes observed in 4D seismic during SAGD (Ma and Leung, 2019; Canbolat and Arturi, 2022)

Deep Basin

Deep basins form adjacent and parallel to a mountain belt due to lithospheric flexure caused by the immense mass created by crustal thickening during an orogeny (Figure 5). The foreland basin receives sediment eroded off the adjacent mountain belt and the craton, filling the basin with thick sedimentary successions that thin away from the mountain belt and the craton (Schulte, 2018).

With the expansion of the West Path Delivery 2023, TC Energy’s proposed plan to modify compressor stations along the Gas Transmission Northwest pipeline in Oregon, Washington and Idaho, and LNG Canada becoming operational, it will put pressure on natural gas operators in the Deep Basin. LNG will also become operational, increasing the demand for natural gas from the Deep Basin, particularly the Montney (Government of Canada, Canada Energy Regulator, 2022; Nickel & Williams, 2023; Quan, 2022).

Goal of implementing AI, predictive and prescriptive analytics and ML into tight sands and shale reservoirs is to improve the resource extraction efficiency per well basis which will reduce the greenhouse gas emission and environmental impacts (Kong, et al., 2021).

Figure 5: Diagram showing the fluid distribution in the Deep Basin, Alberta. Notice that gas is under oil and water, which is counter intuitive. The Deep Basin tends to be overpressured, and there is hydraulic isolation, with all the porosity filled with hydrocarbons. Modified after Masters (1984); Burnie et al. (2008); Zaitlin and Moslow (2006). Taken from Schulte (2018).

Predictive analysis can be used to (Schultz, 2021):

  1. Predict microseismic during fracking to estimate where the fractures will go and reduce induced seismicity
  2. Evaluate acquisition and divesture opportunities
  3. Formulate land acquisition strategies, including amounts to be bid
  4. Reduce exploration risk
  5. Control of drilling and completions costs
  6. Improve production operations.

The projects that AI, predictive or prescriptive analytics and ML were used in the deep basin were:

  1. Artificial neural network to predict the ratio of the production profile to the type well curve in the Duvernay formation (Bowie, 2018; Kong et al., 2021)
  2. Production of the liquid-rich Duvernay formation was modeled using a comprehensive data set including geology, drilling, completion, and production from over 500 wells using Shapley values to select features and machine learning (Kong et al., 2021)
  3. Deep Learning models were developed to predict early production in the Montney (Kim et al., 2020).


The implementation of AI, predictive and prescriptive analytics and ML is not an easy task and requires significant changes within an organization, which include change management and the creation of a Chief Transformation Officer on the board who is an extension of the CEO.

It will also require learning and development. To conduct this learning and development, companies need to collaboratively learn from others across oil and gas and other industries that have accomplished this.

With oil and gas companies moving more towards data science, universities should look at how they can combine AI, analytics, ML, modelling and visualization with engineering and geoscience. It actually can cause more students to become interested in oil and gas. Oil and gas companies should be looking at building a talent pipeline with institutions that develop such programs.

We need a relationship between industry and academia so that students graduating are prepared to step into an asset team and be productive. This is why Rick Warters and I gathered support for an internship program for the Faculty of Science at the University of Calgary to become a reality. It was a great moment in my career to do this.

As data science is implemented in our work, we need to adopt Steve Jobs’ approach to hiring and managing people:

“It doesn’t make sense to hire smart people and then tell them what to do; hire smart people so they can tell us what to do.”

– Steve Jobs

With this transformation, we need to admit what is known and what isn’t known and to hire people to fill that gap. Sometimes leaders need to trust because it is hard to know everything. Just as much as the team is on a learning curve, the leadership is as well. Employees need to realize it is not leadership’s role to know everything, leadership is there to remove hurdles along the way.

To speed up the learning curve, we need to have brainstorming sessions to build transparent relationships across the team, because employees need to feel valued during a transition. Part of feeling valued is in the psychological contract between employees and leadership. When employees feel the psychological contract is not being satisfied, they leave.

The adoption of AI, predictive or prescriptive analytics or ML will involve jumping to the next curve, just like what happened when we went from horse-drawn carriages to the car, or telephones to the internet, or personal computers in our homes.

With that jump, there will also be intrapreneurship within organizations which will involve the lean type of rollout where a prototype will be developed, rolled out so feedback can be given and with that feedback a new prototype will be built and rolled out again. It will take time, it will not be perfect, and we need to adopt what Guy Kawasaki said in a TED talk, “Don’t worry, be crappy” (Kawasaki, 2014). It will take time, but when we jump to that next curve, we will have disruption, which will affect how we do business. This is why psychological safety is important.

Not being perfect causes some to feel uncomfortable because there is this overlying belief that what we do needs to be perfect, which has created pushback in the past when other disrupted technologies, like when AVO was implemented within the oil and gas industry. There was always someone who failed with AVO, and it was why Rose and Associates created their consortium to produce how to de-risk the AVO utilizing multiple attributes such as matching structure, interval velocity drop, positive responses from different AVO attributes, signatures on gathers, well modelling, etc. AI, predictive or prescriptive analytics or ML incorporates AVO and seismic inversion products within their analysis, which is a multivariate analysis using cross-discipline data.

The leaders of any organization should strive for their company to be one of the companies on the “Fortune’s 100 Best Companies to Work For” because these companies tend to be the best-performing organizations on the S&P 500 (Goenner, 2008), and they attract much-needed talent. Employees are looking for a culture where everyone can bring their full range of talents, feel valued, and do magnificent work (Mahony, 2022). During a transformation such as digitalization, it can be difficult because there will be large hurdles to overcome. This is when leadership is evaluated and why the CTO needs to be part of the executive leadership team and function as an extension of the CEO.


Abbas, T. (2023). Beckhard and Harris Change Formula Explained. CMI.,achieving%20the%20vision%20R%20%3D%20Resistance%20towards%20change

Ansari A.; Heras M.; Nones J.; Mohammadpoor M.; Torabi F. (2020). Predicting the performance of steam-assisted gravity drainage (SAGD) method utilizing artificial neural network (ANN). Petroleum 2020, 6, 368–374. 10.1016/j.petlm.2019.04.001.

Beckhard, R. and Harris, R. T. (1987) Organizational Transitions: Managing Complex Change, 2nd Edition. Boston, MA: Addison-Wesley.

Blouin, M., and Bloaguen, E. (2017). Colored inversion.

Boudway, I. (2022, September 20). More Than Half of US Car Sales Will Be Electric by 2030.

Bowie, B. (2018). Machine learning applied to optimize Duvernay well performance. SPE Canada Unconventional Resources Conference, Society of Petroleum Engineers.

Burnie, S.W., Maini, B., Palmer, B.R., and Rakhit, K., 2008. Experimental and empirical observations supporting a capillary model involving gas generation, migration, and seal leakage for the origin and occurrence of regional gasifers. In: Cumella, S.P., Shanley, K.W., Camp, W.K. (Eds.), Understanding, Exploring, and Developing Tight Gas Sands, 2005 Vail Hedberg Conference, AAPG Hedberg Series, vol. 3, pp. 29–48.

Canbolat, S. and Arturi, E. (2022 June 7), Machine-Learning Approach for Forecasting Steam-Assisted Gravity-Drainage Performance in the Presence of Noncondensable Gases. ACS Omega, 7(24), 21119–21130.

Cann, G. (2020). 8 killer ways to apply analytics in oil and gas. Geoffrey Cann.

Correa, O. C. (n.d.). 4 reasons you should invest in Digital Transformation.

Coursera. (2023). What Is Change Management and How to Use It Effectively. Coursera.

David, D. W. and Farzan, S. (2021, August 30). Boards Are Undergoing Their Own Digital Transformation. Harvard Business Review.

DeepAI. (2020). Neural Network. DeepAI.

Derkach. A., Fantaguzzi, I., Pearse, N. and Smith, M. (2023 February, 3). Powering up new leadership for a changing energy environment. McKinsey & Company.

Digit. (2021 April 21). How long can oil & gas keep putting off digitalisation? Energy Voice.

Dion, Martin. (2020). Can AI be a force-multiplier for Intelligence Analysis? 10.13140/RG.2.2.33630.05448.

Farmer, D. (2021). Descriptive vs. prescriptive vs. predictive analytics explained. Business Analytics.

Fu, D. (2018). Unlocking Unconventional Reservoirs With Data Analytics, Machine Learning, and Artificial Intelligence. JPT.

Gartner’s Hype Cycle Methodology – CIO Wiki. (n.d.).

Gartner, Inc. (2021, August 23). Gartner Identifies Key Emerging Technologies Spurring Innovation Through Trust, Growth, and Change.,next%20five%20to%2010%20years%20%28see%20Figure%201%29.

Ghadban, W. (2023 March 10). The great AI debate in the oil and gas industry: balancing job loss concerns with the thriving potential of AI applications.

Goehring & Rozencwag. (2019 November 7). Declining Tier 1 Acreage and Slowing US Shale Oil Production.

Goenner, C. (2008). Investing in Fortune’s 100 Best Companies to Work for in America.’s_100_Best_Companies_to_Work_for_in_America

Gorter, O., Hudson, R., and Scott, J. (n.d.). The role of the chief transformation officer. (1 C.E., January 1). McKinsey & Company.

Government of Canada, Canada Energy Regulator. (2022, May 24). CER – Decision in Brief: West Path Delivery 2023 Project.

Guevara J. L., Patel R. G., Trivedi J. J. (2018). Optimization of steam injection for heavy oil reservoirs using reinforcement learning. In Proceedings of SPE International Heavy Oil Conference and Exhibition, Kuwait City, Kuwait, December 10–12, 2018; Society of Petroleum Engineers: 2018; 10.2118/193769-MS.

Guevara J. L. and Trivedi J. J. (2022). Towards a machine learning based dynamic surrogate modeling and optimization of steam injection policy in SAGD. In Proceedings of SPE Western Regional Meeting, Bakersfield, California, April 26–28, 2022; Society of Petroleum Engineers: 2022; 10.2118/209245-MS.

Kaplan, L. (2023). Canadian upstream oil sector supply costs continue to decline.

Kim, M., Nguyen-Le, V,. Shin, H., Lee, H.S. (2020). Deep Learning-Based Prediction of the Cumulative Gas Production of the Montney Formation, Canada. GeoConvention.

Kolaczkowski, M., Caruso, P., Elizondo, D. (2021 July 5). What new strategic imperatives in the oil and gas sector mean for the energy transition. World Economic Forum.

Kong, B., Chen, Z., Chen, S, Qin, T. (2021). Machine learning-assisted production data analysis in liquid-rich Duvernay Formation. Journal of Petroleum Science and Engineering, Volume 200, 108377, ISSN 0920-4105.

Kumar A., Warren G., Joslin K., Abraham A., Close J. (2020). Steam allocation optimization in full field multi-pad SAGD Reservoir. In Proceedings of SPE Canada Heavy Oil Conference, September 29–October 2, 2020; Society of Petroleum Engineers: 2020; 10.2118/199907-MS.

Lancaster, S., and Whitconbe, D. (2020). Fast‐track ‘coloured’ inversion.

Lasopamenu. (n.d.). Gartner Hype Cycle AI.

Lee, Y.N. (2017 December 18). Robots ‘are here to give us a promotion,’ not take away jobs, Gartner says.

Liang, J., Ramamauskas, B., and Kurenkov, A. (2019 February 19). Job Loss Due To AI — How Bad Is It Going To Be?,in%20general%20will%20likely%20outnumber%20the%20jobs%20displaced.

Luo, G., Yao, T., Bychina, M., Ehlig-Economides, C. (2018). Production optimization using machine learning in Bakken shale. Unconventional Resources Technology Conference, Society of Exploration Geophysicists, American Association of Petroleum Geologists, Society of Petroleum Engineers, Houston, Texas (2018), pp. 2174-2197

Ma Z. and Leung J. Y. (2019). Integration of deep learning and data analytics for SAGD temperature and production analysis. In Proceedings of SPE Reservoir Simulation Conference, Galveston, Texas, April 10–11, 2019; Society of Petroleum Engineers: 2019; 10.2118/193829-MS.

Mahony, M. (2022 January 6). How to Attract Top Talent in 2022 Harvard Business Review.

Masters, J.A. (1984). Lower Cretaceous oil and gas in western Canada. In: Masters, J.A. (ed) Elm-worth – Case Study of a Deep Basin Gas Field: AAPG Memoir, 38, 1 – 35.

McClure, B. (2021 December 29). Can a Company Have Too Much Cash?

McKinsey & Company. (2016 April 21). The need to lead in data and analytics.

Middleton, R.S., Gupta, R., Hyman, J.D., Viswanathan, H.S. (2017 May 9). The shale gas revolution: Barriers, sustainability, and emerging opportunities. Applied Energy 199 88-95.

Mohaghegh, S.D. (2005). Recent developments in application of artificial intelligence in petroleum engineering. J. Petrol. Technol., 57 (4), pp. 86-91

Mohaghegh, S.D., Modavi, A., Hafez, H., Haajizadeh, M., Guruswamy, S. (2009). Development of surrogate reservoir model (SRM) for fast track analysis of a complex reservoir. Int. J. Oil Gas Coal Technol., 2 (1), pp. 2-23

Nickel, R., & Williams, N. (2023, January 18). Natural gas, not electricity, to power Canada’s first LNG plant, increasing carbon footprint. CBC.

Nickel, R. (2021 June 28). Old, small and CO2-intense: why Canada’s highest-carbon oil sites keep pumping.,goal%20to%20cut%20emissions%20by%2040-45%25%20by%202030.

Nilsen, E. (2022 October 10). Biden is blamed for downturn in new oil drilling, but fossil fuel companies are the ones hitting pause.

Orland, K. and Morgan, G. (2023). Trans Mountain pipeline expansion to boost Canadian oil production.

Quan, H. (2022, April 22). Update: Three West Coast LNG projects make progress – Context Magazine by CAPP. Context Magazine by CAPP.,through%20the%20regulatory%20approvals%20process.

Ross, C.P. (2016). Improving Resolution and Clarity With Neural Networks. American Association of Petroleum Geologists (AAPG) Explorer.

Rystad Energy. (2021, November 17). Breakeven Price For New Oil Projects Drops In 2021.

Saucier, H. (2020, April 30). Geoscience Programs Evolve Through Declining Enrollment. AAPG EXPLORER.

Schulte, B.Wm. (2018). Changes in the Oil & Gas Industry – Unconventional Plays (Engineering Fused with Geoscience).

Schulte, B.Wm., Bridge, D., and Lyatsky, H. (2019). Methods of fault detection with geophysical data and surface geology.

Schulte, B.Wm. (2022a). Keynote address Reservoir Characterization – 16 years ago, today and tomorrow. GeoConvention.

Schulte, B.Wm. (2022b). Review of global warming, climate change, renewables, CCUS, geothermal and the geoscience involved: Part 3.

Schulte, B.Wm., Gray, D., and Wen, R. (March, 2023). AVO, Seismic Inversion, Machine Learning and Predictive Modelling and the Economics of the Play – New Age for Oil and Gas Exploration.

Schultz, Y. (2021). Column: More oil & gas predictive analytics applications. BOE Report.

Seeking Alpha. (2014 August 24). Oil Sands Vs. Shale – Who Gets Hit First?

Slav, I. (2022, February 7). Running Out Of Sweet Spots Shale Growth May Not Materialize.

Tableau. (n.d.). Artificial intelligence (AI) versus machine learning (ML) versus predictive analytics: Key differences.

The Canadian Press, (2022 December 22). Canada moves to mandate electric vehicle sales starting in 2026.

The Investopedia Team. (2022). How strong are the barriers to entry in the oil and gas sector? Investopedia.

Toppe, J., (2022 December 30). S&P energy sector’s record 2022 performance ‘built on downfall of others.’ FOXBusiness.

Valle, S. and Kumar, A. (2023, April 28). Exxon, Chevron split over how to manage rising cash piles. Reuters.

Vector Solutions. (2022, June 28). What Is a Learning Organization? | Learning Organization Theory.

Venables, M. (2018, June 17). Oil And Gas Executives Bemoan The Pace Of The Digital Transformation. Forbes.

Wachtmeister, H. and Höök, M. (2020). Investment and production dynamics of conventional oil and unconventional tight oil: Implications for oil markets and climate strategies.

Wang,S., Chen, Z., Chen, S. (2019). Applicability of deep neural networks on Production forecasting in Bakken shale reservoirs. J. Petrol. Sci. Eng., 179, pp. 112-125

Wikipedia contributors. (2023). Gartner hype cycle. Wikipedia.

Williams, N. (2023, January 23). MEG Energy expects Trans Mountain pipeline to be fully operational by early 2024. Financialpost.,300%2C000%20bpd%20currently%20once%20the%20expansion%20is%20complete.

Worland, J. (2022 March 10). The Real Reason Big Oil Won’t Save the U.S. from High Gas Prices.

Wu, S., 2020, Multicollinearity in Regression. Towards Data Science,

Zaitlin, B, A., and Moslow, T. F., 2006, A Review of Deep Basin Gas Reservoirs of the Western Canada Sedimentary Basin. 2006 Rocky Mountain Natural Gas Geology and Resource Conference.

Zhou, Q., Kleit,A., Wang, J., Dilmore, R. (2014). Evaluating gas production performances in marcellus using data mining technologies. Unconventional Resources Technology Conference, Society of Exploration Geophysicists, American Association of Petroleum Geologists, Society of Petroleum Engineers, Denver, Colorado, pp. 20-36.

About the Author

Brian Wm. Schulte attended The University of Calgary, graduating in 1989 with a Bachelor of Science in Geology and a minor in Geophysics. Brian has worked in seismic processing, acquisition, interpretation, rock physics, and petrophysics. Some of the companies he worked for are Gale-Horizon, Schlumberger, Vastar (a division of Arco), BP, Explora Seismic Processing (ESP), Geokinetics, Talisman Energy Inc., and Repsol. Brian also served as an Instructor of Petroleum Engineering Technology at Houston Community College-NE Energy Institute. He made outstanding contributions as a member of the Program Industry Advisory Committee that led to several program recognitions and students’ successes. Brian is working at his own consulting company Schiefer Reservoir and is now attending Sprott School of Business – Carleton University obtaining an MBA in Business Analytics. Brian has served as the Chief Editor of the Recorder from 2018 to 2023.