Extracting more from data: What generative AI means for energy companies

Generative AI holds immense potential for reinventing energy companies. Why? Just think about all those vast reserves of data waiting to be tapped across the industry.

Energy has always been a data rich business. And generative AI’s ability to process, query, analyze and summarize that data near instantly is going to transform not only the way energy companies manage their information, but also how they operate, how they augment their workers, and how they deliver value to customers.

Whether it’s accelerating field development planning or augmenting the operations field force, generative AI enables the industry to dig deeper into its data and extract valuable insights in a faster, easier, more accessible, and, above all, human-like way.

So it’s no surprise that a massive 87% of energy executives say AI is inspiring their long term strategy according to Accenture’s Technology Vision research. Even more — 95% — are either very or extremely inspired by the new capabilities offered by large language models (the technology underpinning generative AI).

This reflects the fact that oil and gas companies have numerous roles where a significant proportion of the working day has the potential to be augmented or automated through generative AI according to Accenture analysis.

This includes 64% of general and operations manager hours, 58% of first line supervisors of retail sales workers, among others. Furthermore, large language models (LLMs) have high potential to automate/augment field worker hours. For example, petroleum engineers (39%) derrick operators (25%) and rotary drill operators (18%).  

An industry ready to capitalize

It’s true there are challenges around data privacy, confidentiality and model accuracy/auditability to be worked out. While it may be acceptable to have a 97% accuracy rate for predicting the optimal placement of a well, for instance, that wouldn’t be enough for recommending the next best-action in a safety-critical activity like a refinery repair. For most use cases, it will remain important to have a human in the loop as validation.

But there are already lots of interesting use cases being explored. And several energy companies have been experimenting with these models for some time, long before ChatGPT lit the touchpaper under global public adoption.

Our research has shown almost two-thirds of energy executives anticipate making significant increases in the resources dedicated to AI in the next three to five years. And we’ve seen recent announcements from Shell about their plans to use generative AI to improve the speed and efficiency of sub-surface imaging.

At Accenture we’ve been working with one oil and gas major to use techniques like cognitive search and semantic modeling, as well as generative AI, to automate knowledge gathering and improve access to data.

A game-changing technology

It’s clear the industry is taking generative AI seriously. But why is it garnering such intense interest?

First and foremost, it’s the value promise and potential to reinvent energy processes. Our research shows over two-thirds of executives expect AI foundation models to deliver accelerated innovation, among other benefits.

Once they begin fully exploring the ways it can intersect with core processes, companies can open up advanced capabilities like complex scenario modeling, knowledge transfer between workers, and even reshaping the capital expenditure cycle.

Another reason the industry is excited about generative AI is the way it democratizes the technology. Whoever you are, whatever your role, you can query a model using simple natural language. And it will respond with meaningful and consumable insights. That takes AI out of the data science lab and puts it into everyone’s hands.

Generative AI is also a significant departure from a technical standpoint. Because these large language models have been pre-trained on internet-scale datasets, they’re already very powerful out of the box. And they can be quickly fine-tuned for specific use cases. That significantly lowers the entry barrier, allowing for much faster adoption.

Early wins

There are opportunities all the way along the energy value chain — not only in reinventing operations and knowledge management but also addressing critical workforce challenges and environmental reporting.

Employee experience will likely be one of the first places to see the impact. Take equipment maintenance for example. Every engineer who’s ever had to repair a pump that was installed decades ago knows just how much documentation you often have to wade through. Generative AI’s ability to plough through that material, summarize it, and accelerate the engineer’s understanding of it, would provide a massive productivity boost.

Similarly, generative AI offers a lifeline for the decades of tacit know-how that risks being lost as an ageing workforce retires. If energy companies can capture that invaluable experience and expertise, the technology will allow them to transfer the knowledge to a new generation of workers and bridge the looming skills gap.

There are also likely to be early use cases around end-to-end greenhouse gas emissions reporting. Because, of course, energy companies are significant consumers of energy, not just producers. Generative AI’s ability to process and analyze the vast data sets  has the potential to support this important challenge.

Longer-term value

The really significant returns will likely come later, however, as foundation models are adapted and fine-tuned with energy companies’ deep wells of industry-specific data. For example, there are interesting use cases around generating new failure modes for equipment or creating capital project optimizations in real time, which require the model to have an understanding of specific data sets like piping and instrumentation diagrams, timeseries data, schedules, and so on.

Another intriguing prospect is accelerating the upfront aspects of capital projects by getting generative AI to create the basic designs, concepts and feasibility studies. Being able to generate an 80% complete design in a matter of minutes would not only reduce project timelines massively, but also enable more accurate forecasts that feed into greater capital discipline.

How to get started

We would always recommend clients approach this initially with an assessment phase. This is about exploring the potential, defining the overall vision and then prioritizing use cases through a deep dive into the intersections with other technologies, the ease of deployment, and the business value on offer.

Then the company can start running experiments with a handful of use cases (in a sandbox environment) and assess the overall business readiness. This can be followed by defining the reference architecture, setting out a deployment roadmap, maybe establishing a center of excellence, and kick-starting a program of upskilling and awareness across the business.

There are new wells of opportunity waiting to be discovered all across the energy value chain, from capital planning through to retail. To find out more about capitalizing on this seismic shift, please add a comment below or get in touch directly.