Leveraging Artifi cial Intelligence in Oil & Gas Industry: Avoiding Pitfalls and Ensuring Success
Dr Sean Otto
Vice President - I oT & Analytics
Cyient Insights

Though Artificial Intelligence (AI) has been a buzzword in several industries in the last few years to improve operational efficiency and quality, Oil & Gas Industry has also marched towards intelligent automation as a means to improve production, reduce operational expenses and maintain high safety standards. However, not all AI deployments are successful and the industry is still learning how to appropriately apply AI models to processes, integrate them into products and, in all honesty, trust a computer to make decisions that humans once made. The article highlights some of the challenges companies face when adopting AI solutions today, whether they build them in-house or not, as well as a few key characteristics that create successful implementations.

Artificial Intelligence (AI) has been a buzzword for several years. Ironically, however, the concept of AI has been around for decades. With the increase in digital solutions worldwide, the inner workings of many products and platforms we use today contain AI models and algorithms, though they are not necessarily advertised as AI. However, as AI increases its status as a game changer in products and services, that trend is changing.

In the oil and gas (O & G) industr y, there is a drive towards intelligent automation as a means to improve production, reduce operational expenses and maintain high safety standards. Through improved sensing technologies, big data storage, cloud computing, and intelligent data modeling, the O & G industry is beginning its own digital renaissance. Despite this surge in intelligence innovation, however, not all AI deployments are successful. We are still learning how to appropriately apply AI models to our processes, integrate them into our products and, in all honesty, trust a computer to make decisions that humans once made. Outlined below are some of the challenges companies face when adopting AI solutions today, whether they build them in-house or not, as well as a few key characteristics that create successful implementations.

To begin, let's review the key distinction between the two types of AI:
  • Generalised Intelligence - what most of us would consider true AI and is often illustrated in movies
  • Specialised/Specific Intelligence - Designed and tuned for a specific activity or set of situations
The second type represents the vast majority of AI being developed and integrated into O & G solutions today. From a business perspective, we can divide AI-driven solutions into four different areas of impact:
  • 'Do more with less' - Improve and/or main tain production
  • Improve safety - Implement safety equipment detection or air contamination monitoring capabilities
  • Risk mitigation and compliance - Monitor and identify real-time issues in operations
  • Provide additional services and solutions - Use generated data to deliver insight, which could reduce operational expenditures
Regardless of the type of solution being built, AI is a very data-driven and technical endeavor. Once an organisation has solved all the technical issues related to gathering data from sensors, communicating that information and storing it either on the premises or in the cloud, then an AI solution can be developed. Keep in mind that resolving technical issues can be more challenging than anticipated and requires continual maintenance and observation. Once these tasks are accomplished, the challenge shifts to creating solutions with that data, which will solve business problems in an automated, self-learning and self-sustaining manner.

Ultimately, good AI is something we experience, not something we do. Doing the legwork necessary to develop successful AI-driven solutions should be the focus of organisations and is where the toughest challenges are solved.

Data Science - the Power behind AI
AI is driven by the discipline of data science. To better understand how to execute and deliver AI solutions, we need to better understand the practice of data science. Data science is an interdisciplinary field that extracts useful knowledge and insight from data and applies it to solving real-world problems. While there is a lot of hype about AI replacing jobs, historically many technologies have minimised or removed repetitive tasks to enable a focus on higher skills (e.g., typewriting pools). AI should not be seen as a "job killer", but rather as a solution that enables humans to make decisions quicker, optimise repetitive tasks and improve safety. Companies will always need physical labor and operational oversight. For example, a pump armed with sensors generates lots of data. However, with AI, that pump could interpret the data and take action, independently of a human operator.

Driving Successful AI projects
Implementing an AI solution enables O & G companies to optimise operations. Locating and managing assets efficiently, effectively and safely is key when adding AI capabilities to any product or service. However, before an AI solution is built and integrated, companies should be aware of the specific activities that actually drive AI success:

Budgeting Long-term: In O & G, the ultimate goal of AI is to improve up-time, increase safety, and reduce operational expenditures. However, that doesn't necessarily mean AI solutions will always be cost -effective, which leads to our first pitfall. Data science is a research and discovery activity, not an innovation activity; therefore, it takes dedicated time, money, failures, and communication. Successful companies approach AI solutions as 'laps around a track' and don't view it as a 'single moon shot'. Unfortunately, this approach creates challenges when budgeting for an ongoing AI project. However, understanding the process and laying out a comprehensive approach will help avoid potential hidden costs.

Educating Stakeholders: AI-driven solutions should be regarded as part of any company's digital transformation efforts but ensuring that all stakeholders are onboard is a challenge every company faces. There is a strong need to educate leadership, management, technicians, equipment suppliers, and finance, operational, management, and engineering teams about the implications and benefits of AI. As previously mentioned, AI must be thought of as an experience and not as a tool. Data science leverages statistical tools to design specific data outputs, but AI is about integration into systems. Because of that subtle distinction, AI solutions impact many different people and departments within an organisation. Companies who start with communication and training are more likely to be successful in their AI efforts, even before they have designed any AI solutions. It's important to educate teams about integration, but don't overpromise on functionality, because AI isn't a silver bullet and doesn't solve all problems. It is a strategic tool and should be leveraged as such to deliver a concerted team success.

Employing External Experts: Another pitfall that organisations may face when integrating AI-driven solutions is the tendency to do everything inhouse. Keep in mind that AI is also a software technology , and effectively implementing new software requires a team with different skillsets. Acquiring those skillsets also can be a challenging task. There are plenty of vendors available that can provide services and components for AI solutions. Businesses need to be strategic about which components they want to own and which fit with their current operations and skillsets. For example, a company that has an IT group focused on enterprise support would be unwise to "dump" software development onto a team that isn't ready to support it. Companies who have been successful with AI-driven connected solutions have placed their software development team within the sales organisation. This strong distinction enables them to build customer-driven solutions without the encumbrance of ensuring the remaining systems are working.

Building the Right Team: Identifying everyone's role is another challenge to tackle. Successful AI implementations are about creating enduring constellations of success, not brilliant shooting stars that ultimately fade. Here are some suggestions for creating a successful AI enablement team:
  • Data scientists create AI models that predict outcomes with a high degree of accuracy.
  • Data engineers enable model inputs and transform model's outputs into high-performing solutions that are scalable and timely.
  • DevOps teams manage software processes and deploy quality software within minutes, not months.
  • UI/UX teams transform model outcomes into something consumable by creating relevant and visual applications.
  • IT teams support the infrastructure need for AI solutions .
  • Data leaders define the right data objectives and align AI-driven solutions with business needs.
If any group overemphasises their contribution to the overall AI solution, it will only result in a project that is less successful. Additionally, data scientists, engineers, software programmers, and business leaders do not always make good data leaders. Data leaders have a unique and diverse set of skills and shouldn't expect to be able to develop them internally. It requires a unique balance of discipline to process while adhering to the chaos of creativity.

The Value of AI
While everyone is actively seeking AI success, the following key elements are critical considerations:
  • Data science without a strategy is in viting failure.
  • Business strategies come first, technology second.
  • Great ideas don't always translate into a viable business improvement.
  • Evangelisse AI initiatives throughout an organisation with internal training and communications.
  • Celebrate the right wins.
  • Just because a great model exists, doesn't mean the business will readily accept it.
  • Failures are successes too, especially when they result in key business lessons.
  • Put an end to silos and allow other departments to join in on the success of AI.
AI is actively changing the O & G industry and driving businesses to be different. Some are ready for that change and some are not. Organisations that are most willing to reach out and partner with vendors will share in the evolving ecosystem, bringing solutions to the marketplace more rapidly. From upstream applications and model permeability changing how we frack shale oil and gas wells, to improving midstream production, and reducing costs as well as downstream impacts on distribution, AI-driven solutions are unlocking amazing value.