News | Blog

The latest Moblize news and updates.

Is tunnel vision clouding our view of AI in Oil and Gas?

08.28.19 | Posted by Joseph Sellers

Often when I’m in meetings with executives, one question is routinely asked. Is AI hype or real? So, I thought I’d share some insights in the hope of generating further discussion.

There is no doubt that everyone in oil and gas especially operators has woken to the fact that all company data is an asset versus a liability.  Thanks to the consumerization of technology, this is a first step in the right direction to realize value from data. This innovation wave has given birth to many new potential opportunities that could have a game-changing impact. However, only time with tell if AI will be remembered in oil and gas history as an “AI bubble” that burst like many past digital transformation initiatives. Every quarter new players in the industry emerge with claims to a magic AI formula to help Big Oil. Only time will tell.

The digital challenges for E&P operator executives are numerous. They have the responsibility to select the right AI partner for their needs across the entire well life cycle when they may not know a lot about successful enterprise AI implementations. Today, AI partners provide solutions for many things: AI to optimize reservoir, drilling, completion, and production, and some partners offer various combinations. I often call this “AI muddle” as none of these AI partners talk to each other resulting in disparate solutions for E&P operators to manage. 

The confusion doesn’t end there. Let’s not forget the other key players in the ecosystem, for example, strategy firms, mid-market business consultants and data science specialized consulting who offer their expertise to build AI solutions in house for the operators or guide them through the digital transition. They have a combination of strategy, operational, and technology frameworks to guide Big Oil through a successful transition*.  With the noise around AI amplifying daily, it’s become increasingly challenging to filter through that noise.  

To start, we should apply learnings from consumer technologies—technologies that we cannot live without today—that have helped wipe out Big Oil from the top five publically traded companies in the world. If we focus on how they apply machine learning (a subset of AI) we can learn a lot. 

These companies focused on simplifying our daily workflows as the key outcome irrespective of technology and making us good busy instead of bad busy. They also focused on building a growth platform, which had salient characteristics.

  • Supported a diverse set of participants and offered opportunities for creating value in many distinct areas
  • Scaled up by accommodating a large user base without adding unacceptable cost and issues
  • Generated increased returns as participation grows
  • Provided incentives for participants to engage regularly and share their learnings
  • Provided development leverage (investment required to build additional functionalities) and interaction leverage (effort and cost for a diverse set of participants to facilitate interaction)
  • Defined practices to guide activities for a large number of participants

With these pursuits, these companies mined not just machine data but also human-generated data. Generally, everyone knows the definition of machine data, but let’s get on the same page with the definition of human-generated data. It’s the videos, pictures, likes, and comments we make. As technology companies mined this data, they started to make recommendations beyond our belief. For example, how does Amazon know our buying patterns and habits so accurately? How does Siri recommend to us what we may not have even thought of based on location? Also, consider focused advertisement through Google.

Not long ago, these companies used to mine only machine data with inaccurate predictions. As the volume of data grew, coupled with the interactions of human-generated data (or what I call “decision context”) grew, the accuracy of predictions grew. Thanks to their clever “growth” platforms, their technology forced us as consumers to share our human-generated data that was previously in disparate systems like emails, instant messengers, and chats.

It baffles me when we realize from our own experiences, how accurate predictions can be when combined with both human and machine-generated data through these consumer platforms.  Why don’t E&P companies think of growth platforms instead of buying into systems that focus only on optimizing machines? Why are we focused on trying to get value from machine-generated data only but let human-generated data such as pictures, videos, and content in emails gather dust on email servers behind a firewall?

Mining machine data is easy as it’s structured, easily accessible and can create value. That’s why we focus on it first. The big question is how will you then incorporate human data into it tomorrow? Will you spend millions later to integrate it? Also, it’s important to ask yourself if the value generated from machine data only is enough or is it more compelling to generate value from mining both machine and human data together.

Let me give you an example of the power of having them both in one seamless growth platform which has the possibility of not only transforming traditional cumbersome operational processes instantly but giving you combined data upfront to mine and create “compelling value” in the future instantly.

Frame 1: Engineer in the office requests video of the next automatic connection   (generated by mining machine data only)

Frame 2: Employee  in the field makes video and posts (human-generated data)

Frame 3: Engineer in the office says much better ROPs in the hotter gamma zones, try to keep DD on point-geo-steering has been pretty calm on target changes so far. After a while, an engineer posts to field Cross Plot saying drilling above the founder point on this run. Need to watch parameters, and requests pictures of bit when it gets to the surface.

Frame 4: Field attaches to the post pics of the BHAs

The industry must think about the bigger picture by learning and drawing parallels from consumer-centric technologies to ensure AI becomes real, not hype. Ultimately, if AI doesn’t prove to create compelling value, the fault is not in AI but us as we may be trying to mine only 50% of data (machine) instead of 100% of data (machine and human).

Note: *I use the term transition, not transformation, as transformation should not take 10 years. 

–Amit Mehta

No Comments