One of the most frequently quoted problems in the media is the shortage of data science and AI capabilities in terms of trained people and practical know-how. But this is changing with more emphasis on data science and analytics education.
The next challenge is finding solvable problems that can actually generate provable value and ROI. Many pilots have been science projects that have not resulted in a demonstrable impact on Profit & Loss. Getting operators actively participating in these initiatives ensures that the objectives are focused on well performance and improvements to the bottom line.
Labeling the data to align the problems and solutions with the years of historic operating data has been a great burden for many operators and engineers. Many operators have detailed logs of all the problems encountered at every well and the decisions made to solve them. They may track workovers, but certainly not for every shutdown and setpoint change.
A solution must be validated to test if it truly works. Understanding of training sets, test sets, cross-validation and other esoteric terms from data science and statistics is needed, too.
Most importantly, buy-in must be gained from operators, so the solution is deployed and operators take action in time to make a difference. This involves technology to deliver the notification in time and confidence that the notification is right.
It all comes down to operators, engineers and data scientists collaborating seamlessly and continuously as a team to meet common goals, even though they are very different people with different backgrounds and objectives, often in different locations and without a shared technical environment.
These challenges have hindered adoption and many large and mid-size producers have been excited by the potential and started investing to build capabilities. But then the implementation reality sets in as they have to scale the solution and prove it works in the field to get buy-in from those used to doing things in a particular way.
It can take a tremendous effort and commitment to change the processes, the systems and the beliefs of an organization. Facing these challenges, the organization loses motivation. According to Harvard Business Review, only 8% of organizations have invested in the core practices that enable widespread adoption. They are not prepared to turn the pilots into lasting success.
Expert Guided Machine Learning is the way to jump over the chasm and breakthrough these limitations to achieve the objectives of the digital oilfield. As the organization starts seeing the benefits at scale without too much effort, motivation rises!
After modeling thousands of wells across different fields, OspreyData has learned how to address critical challenges in AI innovation. To learn more about expert-guided machine learning, click here.