The world’s largest sovereign wealth fund, Norges Bank Investment Management, is leveraging agentic AI to reshape its trading functions, boost productivity and create a new approach hiring talent across its investment operations. Speaking to Trader TV, Nick Dunbar, editor of Global Trading, a Markets Media company, has been reporting on NBIM’s AI journey and spoke to its quant team on the firm’s early move to the to cloud and its in-sourced data and partnership with Anthropic, to in a bid to enable AI agents to unlock institutional knowledge at scale. The result is a more collaborative trading model and one that could spell the end of the traditional “hero trader” among many buy sides says Dunbar.
To find the original reporting and article go to https://www.globaltrading.net/
Interview
Josephine Gallagher – Welcome, Trader TV. I’m here, and I’m joined by Nick Dunbar, my colleague from Markets Media and editor of Global Trading, and we’re here to discuss the top story from Markets Media this week. Nick, thank you very much for joining us from sunny Spain.
Nick Dunbar – Thank you, Jo.
Josephine Gallagher – So, for this segment, I’d like to discuss your story on Norges Bank investment management and how they are using AI to reshape their trading functions, their hiring practices, and how this is also a broader shift across financial market firms and how they build technologies. So, to start us off, you spoke to the head of quantitative trading and the wider quant team from Nordia Bank Investment Management. So tell us, what did you learn from your reporting?
Nick Dunbar – Yeah, so I spoke to Peder Viervoll , who’s the head of quant trading at Norges Bank, and his colleague Eric Dengerud. So yeah, for background, Norges Bank is like the world’s biggest sovereign wealth fund, $2 trillion, and their trading activity is enormous. They spend something like $1.6 billion a year on trading costs in direct and direct trading costs, and that’s a big challenge for them to try to bend that curve and reduce their (cost) footprint in the market. And that they’re doing that for the Norwegian people. That’s their client. Their single client is the Norwegian. What they told me was how they have used AI in the last few years, which is in a very creative and forward-looking way, and is what’s called agentic AI now. It’s basically taking your knowledge and building AI agents that can do boring things or time-consuming things that you used to have to do yourself. So they explained how they did that, and it involves a number of interesting steps. Using technology very much dependent on Anthropic, that’s their big partner, and they put these steps in place, and they’re getting enormous productivity gains, enormous efficiencies across the firm. And I just spoke to the trading function where this is being done directly in that part.
Josephine Gallagher – You mentioned productivity gains there, and we are hearing a lot about this productivity leap across the industry in the use of AI, gives us taster as to how Norges Bank Investment Management and other buy side firms are reshaping their approach to technology builds?
Nick Dunbar – So Norges Bank started several years ago. They went onto the cloud. They built and put their databases onto a cloud system. They in-sourced all of their data, which is interesting. They used to have external providers. They now have them all in sourced. Because databases are just collections of data, and you have to be able to query the data. You have to ask questions of the data, and this is something that is a coding problem to find out how to get answers from your data, understand patterns like in trading data and things like that. So they they’ve been doing that manually, and the sort of leap forward they made was once you connect to a large language model, something that Anthropic invented for its Claude model called a Model Context Protocol or MCP, and you use that as a way of a large language model talking to huge databases, and because large language models sort of know a bit of everything, they’re not context specific. You have to then take the knowledge inside your team, and you put that in what’s called a skill file in Claude, and then those two things, the skill file and the model context protocol, allow you to use the ALLM agent to talk to this giant database and speed things up a lot. So that’s a big leap they made, and they did it before vendors like Bloomberg and LSEG announced their own ventures in that space. So they were very much ahead of the industry in that.
Josephine Gallagher – In your piece, you also mentioned this ” hero trader” or the champion of the trading desk, and how that might become a thing of the past. Is AI changing the profile of the modern trader?
Nick Dunbar – Yes, that’s exactly what seems to be happening in the buy side. I’d say not the sell side because there you have the sales trader, it’s commission based, still a role for them there. But on the buy side, what we see and Norges Bank is a good example of that is the lowering of barriers to communication means that the hero trader just loses that kind of special status. Once you have the LLM, the Claude, or the models’ ability to take that knowledge from one person and transfer it elsewhere, the floor gets raised for everyone, and that means that different traders, like a low touch, high touch traders, they can they can exchange knowledge much more easily, and they can take that that file from Claude, and they can start their own session, and they can ask the database using that file questions that the hero trader of the past might have just kept to themselves. So that’s for me is a is a big change that seems to be happening.
Josephine Gallagher – Understood. Another thing that we’re hearing across the industry is differentiators in the adoption of AI, and one thing that Norges Bank Investment Management is doing is that they had, like you said, this this early adoption of cloud, and that is a big part of their AI journey. So, what do you see from your reporting as the kind of big differentiators in this AI adoption, is it better data or that technical foundation such as cloud?
Nick Dunbar – So, I’ve learned it’s really important to get this right. So, what Norges does, they use a platform called Snowflake, which has become very popular as an AI-friendly database platform, and that’s really got to be the foundation to use these things intelligently. Because you may have like a lot of different queries of a database scattered through your organization, but you’ve got to bring that together and be able to deploy those, you know, with an LLM help. Building that foundation is key. So there are, you know, Databricks, Snowflake. There are these platforms now that are playing a big role in enabling the likes of Norges Bank to do this.
Josephine Gallagher – Really, really interesting. So it’s that foundation layer that can help massively in that AI journey, and so Nick, we could talk for hours about this subject matter. It’s really, really fascinating. But for anybody that wants to learn more about this particular use case and other coverage, where can they find your reporting?
Nick Dunbar – It’s on G lobal Trading, go to globaltrading.net. It’s on the homepage. It’s the lead article, and you have to register, but don’t let it put you off, it’s free. It’s just there to go through that registration process, and you can read it.
Josephine Gallagher – Excellent. Thank you again for joining us on your holiday, Nick. It was great to have you on.
Nick Dunbar – Thank you.
Josephine Gallagher – I’d like to thank Nick for his insight, and of course, you for watching. This has been Trader TV.
