TAKEAWAYS FROM FUTUREAI HOUSTON

AI is now firmly embedded in the energy industry ecosystem

The energy industry is learning fast how to scale up pioneering AI projects, turning them into systemwide applications that speed up work and create financial efficiencies.  

“Let’s remember how much progress we’ve made with AI.” Haavard Oestensen, CCO & EVP, Kongsberg Digital, told the conference. 

“We’re now no longer talking about small use cases, with very limited funding for a single asset, we are talking about industrial scale. Thousands of users are saying these technologies are now impacting how they work and that they need more funding. We should start inviting more CFOs to these conferences to see the value in AI adoption,” he said.

AI should be driving outcomes, not just productivity gains.

“AI is already improving workforce productivity and the user experience, but the bigger value comes when it’s embedded into core workflows to deliver tangible business outcomes,” Justin Lo, Senior Technical Manager, Enterprise AI, at Chevron 

told delegates in a keynote speech.

AI and LLMs are revolutionising data analysis

AI is making everyone smarter.  

“The ability to interact in natural language with your data – to say plot this time series, zoom into this or whatever you need to do – will make everyone a much more effective analyst – and it will probably mean we won’t need to use static dashboards in five or 10 years’ time,” the head of one AI-powered knowledge platform for energy companies told the event.   

The business should be driving AI, not vice versa

Make sure your AI applications are more than just sophisticated toys.

As Ahmed Farag, Head of Business Development, Oil, Gas & Chemicals, Gecko Robotics, put it: “You can’t get to the pyramids in a Lamborghini – it’s not going to take you through those sand dunes. You might have amazing software, but if the software is not designed for the workflow in hand, it’s not going to help you achieve your mission.” 

Kongsberg Digital’s Haavard Oestensen agreed.

“Don’t run an AI project as a centralised R&D initiative with little connectivity to the rest of the kingdom. Ensure that AI initiatives are very much sourced from the business. Involve the business in selecting the problems to be solved – and have people validate the solution and its outcomes,” he said. 

You need to bring the workforce on the AI journey

Gaining trust in AI and getting the most out of it means involving the whole workforce, according to Chevron’s Justin Lo.

“It’s been great to see our leaders link AI directly to financial performance in our most important assets. We also know that sustainable impact requires scaling thoughtfully, bringing the entire corporation along so AI drives efficiency and bottom-line results everywhere,” he told the Future AI event.

A focus on change management is crucial

A culture of enabled change management is a huge competitive advantage. You need to adopt, adapt and learn quicky and then and reiterate if you don’t want to get left behind by your competitors.

“We need to invest more in change management. When trying to solve a problem, get a quick win, get on base to make sure that little win helps, then get another one and then it starts snowballing,” the leader of one energy consulting said.   

Actions derived from AI must be explained – and be open to criticism

AI decisions need to be understood not just acted on. 

“There has to be explainability. Operators who are going to use an AI outcome should understand not just the what, but also the why behind it. Then, the escalation process has to be clearly defined, and you need an audit of how the decision was arrived at and who is taking the resulting actions,” Varun Anand, Enterprise Architect, Pemex, said. 

And a culture of criticism is healthy – workers should be able to question AI output, Riad Efendi, PhD, Global HSE Manager, TechnipFMC, told delegates. 

“AI can make mistakes, even if it is summarizing information verified by dozens of safety management professionals. Because of that, front-end employees should know they have the right, and even the obligation, to look at these suggestions critically,” he said

Take basic steps to make sure AI projects don’t fail

There has been a lot of talk recently over the high failure rate of early-stage AI adoption. To avoid failure, subject matter experts need to be involved right at the start in the development of AI models, so those models are more realistic and grounded, the event heard.

“You also need to make sure you have the right infrastructure and governance in place so you know you’re integrating data quality and other elements in a manner that’s cost effective and repeatable,” a director at a large global IT infrastructure services firm said.

Security needs to keep up with rapid AI evolution

Agility is essential to ensure that data remains secure in a fast-evolving AI landscape, according to Krishnan Sankaranarayanan, Director, Digital Transformation, SABIC.

“Rules relating to data security need to be agile and adapt faster, because the ecosystem we operate in is moving at breakneck speed and they are not keeping up. Every day, we need to make the guardrails higher or lower,” he said.  

Build on existing safety frameworks to speed up AI adoption

“We should utilize and extend the various many process safety management and other critical frameworks we already have, for AI, and not replace them,” Harvinder Singh, Staff Enterprise Tech Engg, Unified Data Platform, bp, told the event. 

“You can classify AI risks in a very similar way to existing process safety management frameworks, from super critical tasks to more mundane, benign tasks,” he said.”