Barely a day passes without hearing about new oil patch use cases for artificial intelligence (AI) and generative AI. These technologies offer benefits like massive time savings when it comes to dealing with data, calculations, and repetitive tasks, and they are becoming more powerful and easier to use by the day. While AI and gen AI can serve as “superchargers” to human capabilities and provide a competitive advantage for businesses, worries linger about the technology’s tendency to “hallucinate” when providing answers and whether AI will replace knowledge workers’ jobs. Even so, those using the tech are optimistic about its potential, although they believe humans should remain involved in the workflows. Speaking during a session focused on generative AI at the Unconventional Resources Technical Conference (URTeC) in Houston in June, Travis Clark, enterprise AI data scientist at Chevron, said the supermajor has applied AI to improve its production in the Permian Basin. “How can AI help extract more oil for less?” he asked. AI is helping the company enhance its shale and tight recovery in a number of ways, including improving well production and frac design, predicting potential frac hits, and optimizing chemical selection to increase recovery, he said. It has also helped the Permian drilling and completions team learn lessons in real time that can be applied to subsequent pads. For example, the team has used AI to create a sand bridging model based on data from the drilling and completion phase to predict the likelihood a well will have issues from flowback so mitigation measures could be put in place. “We’re sitting on quite a bit of data that we need to make sure we’re taking advantage of,” Clark said. Chevron is also taking advantage of the potential generative AI offers. Clark said that while Chevron went “all in” on Microsoft’s Copilot when that became available, one of the early generative AI efforts the company carried out was a look-back project of wells. It summarized data from drilling, completion, and operations logs for the engineers so “lessons learned are carried forward, but problems aren’t repeat problems.” A year ago, he said, there was some uncertainty about whether generative AI could reliably parse unstructured data—which doesn’t follow a predefined structure and is hard to analyze with traditional methods—and then insert it into structured models. Today, those efforts are yielding good results. Chevron is also working to improve a chatbot intended to assist oil and gas operations. A multi-agent workflow with access to engineering and safety documents, the chatbot is a modular and scalable AI-powered assistant that can help surface the right documents and standards for team members, he said. Clark later told JPT that the focus “moving forward is to expand our application of this tool to other areas of our business as we discover potential ways to improve it.” Above all, he said, the company is focused on governing employee use of AI and has established proper guardrails to ensure it is used safely, ethically, and cost-effectively.
Jennifer Pallanich (Mon,) studied this question.