While some in the market may be trying to keep their AI practices quiet, speakers at this year’s ICMA conference in London were proud to highlight their work in the space. Even so, as use cases advance, so does the need for compliance.
“We are probably at the start of our AI journey,” said Louise Drummond, global head of investment execution at Aberdeen Investments. She explained that the firm has built agents to take away heavy-lift tasks for the team, with the agent able to access the company’s data, processes and policies. This has allowed for data analysis to accelerate.
Aberdeen adopted no-touch trading in 2023, she added, primarily on their passive book and low value trades. It has then used AI to analyse the results from this, and is applying the findings to other funds.
“From a trading perspective, that has given us a very good idea of where we think we can adopt that automation next. Putting AI over on top of it has given us the results a lot quicker and confirmed what we thought.”
At BNP Paribas Asset Management, Yannig Loyer, global head of markets and liquidity solutions, was optimistic about the future of AI in fixed income trading. “I think at some point in the future we’ll see some part of the bond markets fully agent traded, from systematic investment decisions to smart order routing to RFQ to liquidity provisions and algorithms,” he said.
“The first use case of AI was parsing runs, which was easy to implement with KYC onboarding and broker selection, and to a certain extent in post-trade, identifying outliers.”
In the primary space, Thorben Lüthge, head of markets and managing director at DekaBank, shared that his firm had introduced a bias-removing AI tool to determine client demand for derivative products and hedging requirements.
“We now have less variance. Every trader on the primary book now gets a recommendation of how much he or she should hedge based on the market parameters, on the structure, on the supply demand situation.”
“It’s actually quite good,” he remarked. “AI is much more neutral than any trader or sales person than we currently have.”
The next step for DekaBank is to implement it in the low-touch flow business, he said, tests for which are already in progress.
“The differentiation [in the near future] will not be who has AI, but who has AI they can govern,” predicted Fahreen Kurji, chief customer intelligence officer at data analytics company Behavox. “Regulators are also getting smarter, and they’re going to care a lot more about how you can evidence and explain with them your models, so that they can actually also control what’s happening in the markets versus just having AI run rampant in the space.”
“We have the same issue with AI that we have with LLMs,” added Loyer. “We cannot let the models make decisions. That means having a certain level of control,” he explained.
The approach here is more about improving the human trader’s behaviour, however, limitations here include the quality and accessibility of data, he noted.
Drummond affirmed the human-in-the-loop approach at Aberdeen: “Everything we do has a trader or compliance or risk involved to make sure that those results are as we’d expect, and that the client outcomes are as expected as well.”
©Markets Media Europe 2025
