RBC eyes AI to bolster FX and rates algos

RBC Capital Markets plans to apply the same artificial intelligence technology used in its equity client execution algorithms to its currencies and rates business, according to a managing director at the firm.

Aiden, as the algo suite is known, is already live for equities, and uses deep reinforcement learning, where an algo learns the best series of actions by trying different approaches and receiving feedback via rewards and penalties. The bank offers an arrival price algo to trade North American stocks and a volume-weighted average price algo in North American and Latin American markets, which can more easily adapt to changes in market regimes.

Speaking on a panel at the FX Invest conference in Boston on April 13, Mike Harris, global head of cross-asset e-sales at RBC, said the bank plans to roll out Aiden in the fixed income and currencies business next.

“We certainly plan on bringing this [technology] into the FX, rates and futures space over the coming years,” said Harris.

I think the underlying logic of looking at 200+ data points and making the [trading] decision at once is something that we can apply to other asset classes

Mike Harris, RBC Capital Markets

He believes the use of AI technology in FX algo execution would be unique: “I don’t think [AI is] being used in FX right now, just to have a selfless plug. RBC does offer AI execution in the equity space. It’s been around since 2016.”

RBC was an early developer of AI technology. In 2016, the bank founded Borealis, an AI research centre based in Toronto with more than 100 staff. In the second half of 2020, the bank launched Aiden, which targeted a number of common client issues, including slippage and changes in volatility environments.

Harris admitted that applying the same logic that had worked for equities to FX and rates would be a challenge: “It’s going to be quite a lift,” he said.

“But I think the underlying logic of looking at 200+ data points and making the [trading] decision at once is something that we can apply to other asset classes.”

Algorithmic execution is already a core strategy of RBC’s fixed income and currencies business, and the bank is one of the few to offer algos to execute US Treasury transactions, sourcing liquidity from the bank’s internal inventory and from public central limit order books.

Other dealers have tried to introduce AI to FX algos but had difficulty making it work.

Speaking on the same panel, Andriy Bukatar, head of algorithmic trading development for equities and FX at Cantor Fitzgerald, said: “Since we went into production with our algos, we were trying to do something with AI. I don’t think we were successful.”

Don Cummings, head of G10 and emerging Asia trading at Mizuho Americas, said it wasn’t clear which, if any, sell-side dealer is using AI algos to execute FX transactions.

“We’ll definitely get there. We’re going to see it—there’s no question about that,” he said.

“I think AI is in our very near future in every aspect of lives, not just rates and FX,” Cummings added.

Update, April 18 2023: This article was updated with information about the algo’s usage in Latin American markets. The algos analyse over 200 data points, not 26 as previously stated.