KUALA LUMPUR (Oct 14): Companies in the oil and gas industry must adopt artificial intelligence (AI) to build resilience against disruptions such as energy transition pressures and geopolitical shifts, says Accenture's managing director for energy in Southeast Asia Lim Yin Sern.

“When you think about new energy systems, which include renewables and clean sources of energy, you have to understand that these are not as constant as our traditional base-load power that’s powered by coal or diesel. You need the ability to flex the power grid, and your energy systems need broader defences, which allow you to plan across these dynamic systems,” he says. 

“By using AI and by leveraging the more connected systems we have in the world today, and with architectures that are being designed to be much more open, allowing for greater integration, we’re now at a point where we’re actually able to build energy systems that are much more dynamic.”

For instance, agentic AI — a system designed to sense, think, act and reflect — enables machines to perceive their environment, plan and execute actions, and continuously learn from the outcomes.

With this, Lim says agentic AI will allow much greater automation that enables engineers to manage larger and more complex systems with fewer human resources. 

“One of the challenges energy companies face today is a shortage of skilled people to manage all of these complex AI models and then take action based on the sensors or alarms, implementing the necessary changes in the plant or system. Having agentic AI systems to automate these actions will allow the same number of engineers or plant managers to control much more in the field. It helps scale improvements more rapidly,” he says. 

Another key advantage is in data preparation. AI agents can harmonise disparate datasets from multiple systems, removing one of the biggest bottlenecks in deploying AI at scale. 

“In a plant, or any energy system, you may be dealing with dozens or even hundreds of disparate systems. Each of these has its own data ontology or language, a different way of organising data. In the past, you would need data engineers to wrangle, clean, and organise that data so it could be used by a single AI model. But today, we’re seeing that AI agents, trained by humans, can look at these different data sets and recognise that they contain similar types of information,” he adds. 

“The agent can then harmonise the data accordingly. So now, you can use AI agents to assist with data preparation, allowing you to get your data to a model-ready state much faster.” 

Ultimately, AI is key to building supply chain resilience amid geopolitical disruptions. This is by being able to create forecasts across different scenarios, says Lim. Organisations can then determine the potential actions to be taken from those forecasts. 

“The power of AI is that it enables us to process a large amount of information and conduct analysis at scale, much faster than we could before. So, when it comes to supply chain resilience, AI is a great example of how we can leverage data, consider many different scenarios, and plan for them, something that may have taken months in the past but can now be done much faster,” says Lim. 

Twenty-five percent of energy executives anticipate a significant increase in AI agent utilisation within three years, with 85% agreeing that AI agents will reinvent digital system construction, according to an Accenture report titled “Technology Vision 2025”.

This means there needs to be workforce readiness and pervasive AI literacy throughout the entire organisation. Adoption rates will be affected if skill gaps are not addressed, says Lim. 

On top of that, the human impact of AI must be considered from the very beginning, he says. This means designing for responsible AI and establishing accountability and governance frameworks from the get-go to address potential fears around job displacement. 

There are five key requirements to do this. These are to establish governance and principles upfront, regularly perform AI risk assessments, regularly perform responsible AI testing, ensure explainability and transparency, and prioritise safety.

“The important thing is that we design for responsible AI from the start. We design for accountability and governance from the beginning. And we also need to ensure that, as we develop AI systems using a responsible AI approach, we are consistently thinking through the human impact of everything we’re doing,” he says. 

“You need to create a culture of 'teach to learn', where you get the humans in the loop, those who are dealing with these issues, to teach the AIs. And through that process, they also learn how to work better with AI and improve the AI systems as part of the broader transformation.”

But how should companies start? Lim says it comes down to identifying the critical problems that need to be solved and begin applying AI to solve those problems in a scalable way. In digital transformation programmes, one of the core mantras is to always begin with the pain point. If companies can understand the magnitude of the problem and the potential benefit of solving it, that will help them to prioritise “because there are always a thousand things that could be done”.

Every transformation programme needs a masterplan, guided by a clear view of value. A typical programme may include anywhere from 50 to 500 initiatives. The ability to project the potential value of each initiative becomes critical in deciding what to prioritise.

“As for the challenges, it's more about how we think about them. There are many different challenges to AI adoption, but the most important thing is that we don’t let perfection be the enemy of progress,” he says. 

“The systems available today will continue to evolve rapidly. There will always be something new on the horizon; you can’t wait for the perfect architecture to appear because it will always keep evolving.”