The industry has finally warmed up to Artificial Intelligence (AI) technologies. This is good news. AI technologies are a great opportunity for operators to increase their efficiency and boost their competitive advantage and safe operations. And, in fact, many have already done just that.
But not every company will deploy AI at the same pace or with the same capabilities. Skills to build meaningful and impactful AI models are still scarce. Also, the people who understand the earth’s and hydrocarbon’s physics are not always the same ones who know how to use “Data Science (DS) and AI” tools. For that reason, some Exploration & Production (E&P) companies opt for a centralized DS team that has skills in statistics and building AI models. In this centralized model, engineers’ “exploratory” and “what-if” analyses go to a central team to build a predictive model for their hypothesis.
This process is not ideal. It takes longer to reach to an acceptable and final model even when a request is prioritized (long cycle time). How can you speed it up? While a central DS team may seem to be the only option at the moment, I would argue that eventually those who understand the physics should be the ones building and confirming their own AI and Machine Learning (ML) models. I argue, some of brightest engineering and geoscience ideas are yet to come out to industry, once they learn the tools.
For some petro-proefessionals, learning a new coding language (python for example) is not fun. After all if they wanted to be coders they would be in a different place. We have been there before. The situation is analogous with the old DOS commands days, which were the domain of specialists and enthusiasts. But as soon as software (like Microsoft Office) came out with Graphical User Interface (GUI) – where we clicked and typed a natural language – then the engineers, geologists and the whole world came onboard. Well, AI and ML tools are making that turn and do not require any coding.
It was no longer a stretch to get petroleum engineers to use Excel spreadsheets, then and it is an expectation now. We should have the same expection for them using advanced analytics, AI and ML tools.
I leave you with some tips to reach that goal (if you agree with the above premises);
- Look for advanced analytics tools with graphical user interfaces (as opposed to those that require you learn a new language like python). I like this software’s interface Data Robot, www.datarobot.com
- Make sure your software can easily connect to data sources – it’s not enough to have a software with good GUI.
- Automate the flow of information and facilitate access to information – this is key to success, especially for data warehouses/ lakes or hubs.
- Make sure your users are trained in advanced analytics principles and on the software. At the very least, engineers should be able to build some basic predictive models.
- Consider cloud data warehouses and tools to get the power to run large datasets such as logs and seismic. We are following Snowflake https://www.snowflake.com and MEME SQL https://www.memsql.com/
Certis consults and delivers data services. We are systems agnostic. We focus on helping oil and gas clients set up their backbone data and processes. Find out more about our services