Shell Talks Generative AI at Web Summit

In an on-stage interview at the Web Summit conference in Lisbon, Portugal, on Tuesday, which was attended by Rigzone, Shell’s Head of Artificial Intelligence, Amy Challen, sat down with Tom Lee-Devlin, the Global Business Correspondent at The Economist, to talk about the company’s experience with generative artificial intelligence (AI).

“We’ve had an AI program for quite some years, and we’ve actually had a generative AI program as well,” Challen revealed in the interview.

“Obviously it wasn’t as big, and the capabilities weren’t there, but for example we’ve been experimenting with large language models for years because we reckon it’s a much, much, better way of leveraging our knowledge assets,” Challen added.

“The way we think about this potential is … it’s huge. I’ve always thought the potential of AI is huge anyway and it tends to be underleveraged in large corporates, just because, you know, we’re not digital natives, you’ve got a huge legacy of analogue things you’ve got to catch up on,” Challen continued.

Looking at the potential of generative AI in the energy industry, Challen highlighted several avenues.

“There are so many processes in any industry … whether it’s all the basic functions like HR, contracting, customer support, anything like that, you can speed up, you can get better access, it’s a better experience for everyone,” Challen said.

“There’s also knowledge management which … we’ve been experimenting with for some time and I think particularly in any industry where there’s a big scientific and technical component, you’ve probably got a load of logs, you’ve got a load of experimental log books, engineering notes, maintenance notes, and they’re probably tucked away somewhere, inaccessible, so you’re not fully leveraging the experience of decades in order to maintain and fix your assets and develop the new fuels and technologies,” Challen added.

“So, we’ve been looking at it from that perspective in particular, how we get better access to things to do it better, faster, and speed the energy transition because developing better fuels and better energy solutions is what we’re trying to do,” the Shell representative continued.

“And then I think there’s a third piece, which is sort of more general R&D … and finally there’s just the personal productivity component, so all the pieces like Microsoft copilot for example, which has just come out on general lease, and that’s one where for everyone with a desk job, you could do things just that bit quicker, find things that bit more easily,” Challen went on to note.

In a section of the interview focusing on the potential risks that come with generative AI, Challen said she thinks of the risks in two broad categories.

“One is genuinely related to technology … There’s external risks like cybersecurity and impersonation, then there’s information leakage because people are using these public tools like ChatGPT, and many companies have actually banned access to these tools because they’re seeing their own IP appearing,” Challen said.

“Then there’s what you do internally and making sure that we protect from bias, hallucination, to all the inaccuracy, data privacy concerns,” Challen added.

“And then there’s another type of risk which is just basically, investing wrong and failing to deliver value. And people maybe focus on that first point, so particularly hallucination, but … I see it in a long history of bad software and bad AI and protecting the risks of those,” Challen continued.

Shell’s head of AI revealed in the interview that Shell often has a human in the link.

“At least until we’re really satisfied that something works well, you always have a human checking,” Challen stated.

“And that’s not as simple as just saying - here’s a recommendation, now human you go and check it, because humans get complacent, they see 99 good answers and they don’t check the 100th because it’s been so good,” Challen added.

“You’ve got to design it into the process so that human is doing meaningful work to genuinely validate ‘is this working’ and another thing you can do is just use other AI … If you do things like fact checking, that’s something you can do with conventional AI very nicely,” Challen continued.

“I think the second type of risk … is just more about how do you properly invest in this and deliver value. Because if you get too excited about anything, you’re not going to invest well,” Challen went on to state.

When asked how Shell has thought about bringing senior leaders, as well as frontliners, on board and making sure they have the knowledge and the skills to make the most of the new technology, Challen responded candidly.

“I think we’re not there yet,” she said.

“I’m not going to claim we’ve educated everyone, we’re totally there, our change management programs are perfect.

Asked if Shell has found it difficult to try and get sufficient expertise around generative AI, Challen said, “it’s always been difficult”.

“It’s always about, whenever you’re working in any large organization the talent development model is fixed for a certain type of career path, and technology career paths are always going to be slightly different and that’s always a battle, and that’s just how it is,” Challen added.

“But actually, we already had a really strong team and we were already experimenting in these kind of areas, so while we have … brought some more people to help us with this, we’ve actually been leveraging the people we had anyway and it’s actually worked really well,” Challen continued.

“We’ve also leveraged our partners in a big way … that’s a really big part of it, particularly when you’re trying to ramp up quickly you do not do it alone. You really prioritize what you want to do internally and you really leverage your partners large and small,” Challen stated.

“I’m talking about Microsoft but I’m also talking about startups we work with really closely,” Challen went on to state.

In a separate section of the interview focusing on working with partners, Shell’s AI lead said she thinks it’s critical.

“Anyone who says they’re doing it all internally, well they’re not because one thing that generative AI has given us, or the excitement around it, it’s given us a lot of basic tools, foundation models, open source models … I guarantee no one will be doing this alone,” Challen stated.

“Everyone else is using someone else’s tools, someone else’s models as well and building on top of that and that’s absolutely what we should be doing, but in addition we have to look really carefully at what use case we want to address and say what’s the best way of doing it,” Challen added.

“We’re an energy company, so we don’t want to invent the latest tax administration software because it’s hugely complicated, takes a lot of investment and someone else will be doing it who better specializes in this, let them do it, when they’re ready, let’s use it … we want to focus on the things we know we can do really well and that’s the bits where we have the data and the IP internally to do it well,” Challen continued.

When asked to what extent she worries that generative AI could crowd out other types of AI, Challen noted that “it’s definitely a danger”.

“Certainly if you look at the excitement  - I don’t want to say hype because I don’t think it is hype I think this can be amazing I really do, I have seen a step change in capability - but the excitement has made a big difference to how we operate,” Challen said.

“So normally we work closely with our businesses and functions and it works both ways. Some of them come to us to ask for a solution, sometimes we go to them, it’s an ongoing conversation, but what we’ve seen since ChatGPT came up is they come to us a lot more,” Challen added.

“And people often come … saying we need generative AI, and were like ‘to do what’, and they’re not sure they just know they need it,” Challen continued.

When asked in the interview if it is a greater risk for companies not to provide generative AI tech than to provide it, the Shell representative said there’s quite a big difference in opinion in how best to address this.

“I think it’s one area where you can see, actually, for a lot of questions you ask in the workplace, it’s fine to use public ChatGPT, for example. It’s fine to say can you give me some advice about how to ask for a pay rise or … give me ten use cases for the energy transition,” Challen added.

“It’s when people start sort of putting in confidential data that you start to worry. So, the key there for me is more about education … you want a workforce that’s able to use this because they’re going to have to use it, they’re going to have to use it in the future to be really effective, but they have to use it knowing what the risks are and knowing how it works,” Challen stated.

“So, they have to know about the feedback loops and they have to know that this is confidential information and they shouldn’t be sharing it. So, I do see as an increased risk but a lot of that isn’t so much a technological solution it’s more like the whole upskilling movement that is going to be, I suppose, it’s going to be the hallmark of the whole generative AI revolution,” Challen continued.

In a section of the interview focusing on data, Challen noted that there’s always a risk that everyone gets too excited about generative AI and forgets about data.

“And the good thing about this is, one, generative AI is actually really good at cleaning up data, unstructured data, but also it’s absolutely critical for making good use of generative AI,” Challen said.

“Some of the core things we want to work on are both that use our data and our IP and when I say data I mean documents, logs, as well as structured data, so it’s actually a really good rallying call to say let’s finally sort out our data and make it properly available for people to properly leverage it,” Challen added.

“Unless you have good training data, you’re never going to get a good model you’re never going to get good outcomes,” Challen went on to state.

In a release sent to Rigzone, Web Summit, which describes itself as the world’s largest tech conference, stated that the 2023 edition of its flagship Lisbon event is hosting 70,236 attendees from 153 countries.

“Web Summit has brought 70,236 attendees from all over the world to the Altice Arena in Lisbon to discuss and debate ideas for the future,” the event organizers noted in the release.

“The cohort of CEOs, startups, investors, media, policymakers and creatives will be taking part in 1,180 investor to startup meetings, 70 masterclasses, 17 PITCH competitions rounds, 25 evening events, and Night Summit parties around the most exciting neighborhoods of the city,” it added.

“Over the last decade, as Web Summit has grown, we’ve brought together tens of thousands of people who have used this week in Lisbon as a springboard to do remarkable things – to launch companies, find investors, unveil projects, advance a vision of the world worth debating,” Katherine Maher, the CEO of Web Summit said in a speech on the opening night of the conference, the release highlighted.

Back in September, Hussein Shel, the Director, Chief Technologist, and Head of Upstream for Energy and Utilities at Amazon Web Services (AWS), told Rigzone that organizations across the oil and gas industry are increasing their adoption of machine learning and artificial intelligence to innovate and address a wide range of use cases, from emissions monitoring to production optimization.

In June, Shel revealed to Rigzone that AWS believes generative AI will have a profound impact across industries.

To contact the author, email andreas.exarheas@rigzone.com