Microsoft on Closing Energy's AI Pilot Gap

Praven Gonabal:

Hello, all. And it's my pleasure to welcome Jon Olav Åbeland, Microsoft's industry expert on energy and AI. And we're having a discussion today about the art of scaling AI in the energy industry, the challenges of moving from pilots to production and harnessing value at scale from AI. Welcome, Jon Olav.

Jon Olav Åbeland:

Thank you very much.

Praven Gonabal:

I think it's fair to say that there's a lot of excitement about AI in all industries, but also definitely in the energy industry. The energy industry, in the past, through multiple waves of technology, has been an early adopter, and there are unique challenges that the energy industry faces when it comes to any technology adoption, and unique opportunities as well. But I think there's a sense in the industry that there is a pilot gap. There are a lot of pilots being launched, but in terms of scaling those pilots into production and then taking those to the ultimate outcome of actually getting value from it, it's not quite successful. So do you share that sense? Do you agree?

Jon Olav Åbeland:

Yeah, I think so. There's always been a gap between a pilot and production. You always have had a need to scale up the compliance, the security, all these different types of things. But those are something that has been embedded in organizations over time now, so that is a gap that everyone has experience in bridging. But with AI, you're introducing new angles of attack to the AI. You're introducing new dimensional things you have to consider. So there is a lot more to test, there is a lot more to scale. So that complexity is something the organizations are working hard on trying to implement. But that does make the gap larger. It makes it slightly more risky. I think a lot of organizations are trying to mature this, and they have a lot of AI projects already in the running, which makes it difficult because they have to do this in parallel.

Praven Gonabal:

Are there fundamental things that people should be doing in an organization in an enterprise context that can help them go about this in a more structured fashion? And when they launch the pilot, they already have the view towards potential future scaling as well. Are there such foundational things that the people could be doing?

Jon Olav Åbeland:

I think scaling the complexity is the difficult part, not the volume part of it. So I think they should start and build a vision for where they want to go, what type of agents they want to have, how they want to prioritize it. So you can use those projects to enable the governance that you need to have, the compliance you need to have, the security setups, and obviously build that experience in the organization before you start to go into the multi-agent, the big, complex AI projects that eventually will come.

Praven Gonabal:

So we know the oil and gas industry is famously drowning in data, right? So there's been decades of data. There's different formats of data, all the way from subsurface data to production data. So what do you feel is a good first step that organizations can do to manage the data better as a foundation and as an input into their AI initiatives?

Jon Olav Åbeland:

I think the data unification that we speak about a lot is something that's been an issue for a very long time. We see in oil and gas that it has been so profitable that they have basically just thrown money at problems for a long time. So there is a backup of technological debt that you need to solve to do AI projects. But it's the same if you want to do analytics or other data projects, or you want to unify data across. So what we're doing a lot of now is working on the ontology of data, which means how is the data connected, how can we make the AI make users understand how the business processes work together so that an AI can get data across structured data, streaming data, and all these type of different data sources that we have. So that is a focus we have a lot of in Microsoft now, supporting the ontology across, which helps both the users and AI and applications on top of it.

Praven Gonabal:

So when it comes to architectural basics, so what do you think organizations should be doing to get their architecture in a good shape to be able to sustain scaling on AI?

Jon Olav Åbeland:

What we've done in Microsoft is we've joined our different tools together in what we call Fabric, where you get the integration of security and different AI tools, data agents and operational agents and these type of things, to simplify that part of it so that you can faster get to the meat of it, which is your domain expertise, how do you bring value to that, how do you get data to your fingertips as fast as possible. So I think that is one way that we are helping from the Microsoft side.

Praven Gonabal:

So when launching an AI initiative, what should organizations think before the initiative in terms of making sure that they're setting up the cybersecurity aspects right, they're thinking about explainability before the fact?

Jon Olav Åbeland:

Yeah. Cybersecurity is many things, many dimensions, right? The attack from outside would be those type of spamming attacks. But there's also the internal security, like logging everything you're doing. You need to log why a decision was made, what data was used, what agent was used, what version of the agent was used, and all these type of things. So that's very important, and I think that's something that should be done from the start. If you don't do this, you will run into a scaling problem, and you will be fighting fires in your data system for a very long time. There are new processes and new things that organizations have to mature on this.

Praven Gonabal:

What do you feel the next 18 months, what do you see the prognosis for AI adoption in energy?

Jon Olav Åbeland:

The thing that will happen in the next 18 months is that companies will have to step up and really focus on these things. The AI wave will just continue. There will be new models, more advanced agentic setups, agents working together, helping you more and more, taking more responsibility, doing more advanced jobs for you. And it is just a future we have to prepare for, and there are no corner-cutting here. But there's still a lot of work that has to be done. So it'll be very interesting, and I think in 18 months a lot of things will happen that we can't even imagine today.

Praven Gonabal:

Indeed. But I think the heartening thing really, Jon Olav, is the foundations or the fundamental aspects remain the same. Take care of your data, take care of your architecture, make sure that you think about cybersecurity and the long-term questions, and the immediate specific challenges or the specific opportunities with AI, and you'll be able to navigate them much more safely and much more effectively.

Jon Olav Åbeland:

Yeah, I think the value of doing things right and the way that you modularize your applications and how you turn them into be agile to easily plug in a new agent or to plug in a new data set and these type of things, the people who can do that, the organizations who can do that will be the winners.

Praven Gonabal:

Thank you, Jon Olav. I appreciate your thoughts on this, and it's going to be an exciting few years ahead. And I think when we look back at this moment in time, it'll feel like the moonshot moment all over again.

Jon Olav Åbeland:

Yeah. Thank you very much.

Praven Gonabal:

Thank you.