Navigating AI opportunities and challenges

MARTINE HANNEVIK     Welcome to the Trust in Industrial AI video series, where we explore how to implement AI with speed and confidence in safety critical industries. In this first episode, we'll start high level and explore some of the new opportunities as well as the some of the new risks and challenges of implementing industrial AI. We will also discuss why it's so important to build safe AI from the beginning and to ensure systems are trusted by people, organizations and societies. I'm your host, Martine, and today I'm joined by Kenneth and Christian. Welcome to both of you.

KENNETH VAREIDE     Thank you, Martine. Great to be here.

CHRISTIAN AGRELL     Yeah, great to be here. Thanks.

MARTINE HANNEVIK     Great. I think we'll start with a big question to you, Kenneth. Why should AI be a key discussion point in every top management groups, especially those working in critical infrastructure industries?

KENNETH VAREIDE     It's a great question and a question that we have a lot with our customers these days and it comes up a lot. And one of the things that when we have this conversation, we reflect on what actually went on with digitalization and the industry 4.0. We remember that ten, fifteen years back. And did it deliver on its promise? And the answer to it is partially. And there were a few things missing. And why is that? And this is particularly for the industry where you have assets in play, maritime, oil and gas, the renewable energy and sort of what I get out of these conversations are that we didn't really transform. We digitized existing processes and we perhaps also underplayed the importance of people in this. It was not only about technology, but also about people. And it ended up being big IT investments with cost overruns and what have you.

So a few things we kind of went missing, but by a by a long while, we also get something out of it. So what makes AI different this time and why does it matter for top leaders? AI kind of changes the game in many respects. It goes much deeper in terms of what you can do with it, in terms of the analytics you can do, but also the level of automation you can do. So it is a game changer in many respects and for our customers and for the people we talk to, this has to be reflected. What is the risk and what is the opportunity space that this presents to us? Yeah, so balancing the risk and reward aspect.

MARTINE HANNEVIK     Yeah. And you, Christian, you've also been in a lot of these customer conversations. And what are some of the specific use cases that are being discussed?

CHRISTIAN AGRELL     Well, there are, there are many examples and we see more and more examples and use cases every day really. And for industrial applications, I think especially maybe optimization is very important. So using AI to then find a new design solution or to operate system optimally. So for instance, in, in energy to predict supply and demand and then be able to utilize the energy system in a smart way using AI is a good use case. Also automation. So AI is a great enabler for autonomy. And we see this in maritime. So there AI is used for situational awareness to kind of be the ears and eyes of the vessel, but it's also used for navigation, for route planning, for control. And when you put this together, then you can achieve high levels of, of autonomy, right? Another very important area is predictive maintenance. So this is where we really want to determine when is a good time to do an inspection or a repair. And then AI is used for, for instance, anomaly detection or you can use models that can predict when a failure is likely to occur, for instance. And then this is, this is really about kind of balancing then, OK, what's the, the cost of doing an inspection and repair versus the risk of not doing it. And I think the, this is interesting because the link between risk and AI is, is interesting because we often say that, OK, if you bring AI into a system, well, then you create new risks. And that's true. But if you really think of it like predicting maintenance, increasing autonomy, at the end of the day, these are things that actually increase safety and reduce risk. So if you think about, you know, these use cases where you use AI on an offshore platform, it will have also a huge positive impact also on the risk and not only on the performance. But I should end also by pointing out that, you know, OK, these are different application areas, but this also, this is what we call kind of special purpose AI solutions or technologies that are being used. And it's a bit different than generative AI language models such as Copilot and ChatCPT, which is also very useful, but maybe more used back-office as part of the business operation. So it's, you know, we have different industrial application areas, but it's also using different types of AI.

MARTINE HANNEVIK     So there are a lot of potential benefits, but how do we ensure that they actually get the return on these investments?

CHRISTIAN AGRELL     Yeah. And that's where we see companies and industries are at different levels in terms of how well it has adopted these new technologies. And many are kind of in the awareness stage, others have started to experiment and doing some pilots and then you see some is starting to put it into production. So it's a big variance here. And of course, it has to be addressed in a proper way. So the earlier you can address the issues, the better it is to tackle them and not wait until you put into the implementation because then things get costly. So managing the risk and understanding the opportunity space, but also the risk in the early phase is essential to really reap the benefit out of it. And that has to be done as you walk through the various stages. I think to add on to that, I think that's right. And that's what we see. And we did a study amongst energy professionals, but we also saw that quite many are now creating AI solutions and piloting and testing this. But there are not that many that are kind of taking it really into operation, right? Because and we see this, well, it's in energy, but also in maritime and healthcare. And these are sectors where the consequences of an AI failure can be quite severe. So taking the step from a prototype and into operation where we really have to trust that it works as intended, that can be a big step, right.

MARTINE HANNEVIK     So actually not trusting is a barrier to the implementation.

CHRISTIAN AGRELL     Yeah.

MARTINE HANNEVIK     Are there other barriers that we typically see?

CHRISTIAN AGRELL     Yeah, there are. And it's basically a combination of barriers, I think. So of course, it's a cost issue with these technologies also now a lot of uncertainty around regulation and the AI regulation, cybersecurity concerns and also like an actual resistance to change. You might need news skills and competences as well. So it's difficult to say kind of in isolation, what are the barriers? Because it's not only about the data, the algorithm, the model, it's also about the people and the processes, right.

MARTINE HANNEVIK     And a bit back to kind of the issue with trust, what are some of the things organisations can do to enhance the trust of these new applications?

CHRISTIAN AGRELL     Well, I think there's a nice saying in AI that you cannot make AI safe. You have to make safe AI. And I think there's a lot of truth to that. And it, it's not only about safety, it's about security and more generally about what we call trustworthy AI. So, OK, one thing is to, to create something, a new piece of technology and then try to make it trustworthy. That's very difficult. So it's, it's much better to try to build trust into the system from the beginning. And how you do that technically, that could maybe be a topic for a different day. But I think if you ask about what organisations can do to really be prepared, I think first of all, making sure that that the AI is a top management priority. I think that's important also to understand, OK, what do you want to achieve with AI really? And, and what then what are the relevant types of AI to consider? Also consider, I think about what are the risks involved and how is that compared and to the potential value creation? And then they have more technical things like the data. Do you have access to the right data with the right quality? Do you have the right cybersecurity controls in place? And also now to really look or monitor the, the development of the AI regulation to ensure compliance. I think that's important. And in the end of the day, also making sure that kind of the, the connection between the technology and people, there's a lot to do also there.

MARTINE HANNEVIK     Absolutely. And those topics are actually the ones we will cover throughout this whole series and deep dive into it. But if we start with the first one, top management priority, what's DNVs top management commitment to AI?

KENNETH VAREIDE     So this is very close to the sort of customers and the industry that DNV is serving. So it naturally become a topic, an important topic for DNV. So, and it's closely connected to of the purpose of DNV, safeguarding life, property and environment. So, and we have for decades been committed to research and serving the industry and its wider purpose with the research that we do. So we early on have invested in really digging deep in our research group to look into AI and we have a large number of researchers that is dedicated to this topic and to understand it and how it actually have implications on the industries that we that we serve. So that's a strong commitment that we have. And obviously that also goes into how we implement AI into our own services and as we are delivering value to our customers, but also help our customers to adopt AI in a safe and responsible and trustworthy manner.

MARTINE HANNEVIK     And if you could give our audience one final advice, what should they, what should they think about? 

KENNETH VAREIDE     I think first and foremost is to be curious and be open minded and really exploring the potential that AI introduces. But then also then being very clear about how you're going to manage the risk and address these issues as early as you can, because then you can really quicker move through the various stages and getting into successful implementation and managing the cost also.

MARTINE HANNEVIK     Yeah, I think a key take away for our audience is definitely make safe AI from the beginning.

KENNETH VAREIDE     Yes, definitely.

MARTINE HANNEVIK     Thank you very much, both of you for these great insights. And thank you to our audience for tuning in. If you have any questions or want to learn more about how DNV can support you with safe application of industrial AI, then please visit our website. Thank you.