FX automation plans focus on predictive analytics – panel

As banks and technology vendors look to further automate foreign exchange workflows, some are exploring artificial intelligence-based tools that can predict client behaviours and amend prices without manual intervention.

Speaking on a webinar hosted by Risk.net’s sister title FX Markets on June 11, panellists debated the future use cases for AI and machine learning technology for analysing client trading patterns and market data.

John Stead, director of sales, enablement and marketing at smartTrade Technologies, explained that banks are looking at using AI solutions to differentiate pricing.

“If you can be ultra-responsive to say, ‘this client has this position [and] the market is volatile, and I want to adjust my pricing profiles or my sales margins on the fly in an automated way’, that’s where you can really make those tiny differences. So, the ability to respond automatically, on almost an individual trade basis, and price it specifically per client, that’s where we’re going. And that has to be done through automation,” said Stead.

Many banks have begun trialling this technology to create trade recommendations, trade summaries and portfolio summaries, while using generative AI and large language models in chats with clients.

Dmitry Ilyaev, global head of FX spot and eFX trading at Commerzbank, said banks have already been using AI to help bridge workflow gaps between different FX instruments.

“Various forms of AI have contributed to workflows on an FX desk already, whether it’s using some form of learning to tune strategy parameters, execution parameters or pricing strategies,” said Ilyaev.

“Now, it’s looking at products that are a little bit more voice driven and trying to bridge the gap between voice liquidity and electronic liquidity, and helping to automate those workflows.”

But when it comes to market-making, dealers are looking at more advanced ways to predict how clients will behave and adjust pricing models accordingly, said smartTrade’s Stead.

The technology could also allow banks to query data sets to guide interactions with clients.

“Having FIX GUIs which just show me a history of rejections, that’s going to be quite old school. Soon it will simply be, ‘tell me what are the problems with this client’ and the platform will come back and tell you, ‘X, Y, Z are the issues you should be looking at, and if you’re going to talk to this client, talk to them about this’,” he said.

“Querying the data using AI is 100% the future, but it’s not for every single client. Some of this stuff is five to 10 years away, but it’s our job to sit and think about how we solve those problems now because the banks will use them in five years’ time.”

All about the data

The future success of these solutions will also depend on the availability and accessibility of the data that feeds into them. 

“It’s tough to build a proper AI algo system around transactions if you don’t have all the right data. And that’s something that I know is a big point of contention around who has the right data in FX. Who is the owner of the real highs and lows of the day and opens and closes on certain non-deliverable forwards markets, and things like that?” said Garth Appelt, head of derivatives, FX and emerging markets macro trading at Mizuho Americas.

Appelt added that AI-based solutions could also provide corporates and other asset owners with tools that can level the playing field with more sophisticated hedge funds that have large budgets devoted to data and analytics.

“Banks like ourselves will be able to give them better tools, but we can’t offer something unless we know that it works. And so that means we need to have perfect data to give the advice and to run the historical analysis on it,” said Appelt.

While predictive analytics tools are being touted as a future solution, others are focusing on ways to break down data silos.

Conor Daly, head of FX solutions strats and head of eFX sales for Europe, the Middle East and Africa at Goldman Sachs, said on the panel that joining up data systems across asset classes will help buy-side firms avoid costly operational risks.

“Either they’re having to transmit and handle data across multiple different systems that are all vertically aligned, or they’re taking data out of systems, manipulating it in Excel, making changes there, then creating execution instructions back in. Every single layer that you add to that calculation process – from raw exposure through to trade, ultimately – is adding operational risk at some point,” said Daly.

“I think joining up those systems much more systematically has to be the focus for both buy-side and sell-side participants in the future. The best way that this is done, thinking about empowering our clients and empowering investors, is in as open-ended-turnkey a format as possible.”

Appelt agreed that joining equity and FX trading systems could also provide the buy side with significant operational efficiencies.

“There are tons of equity trading systems, there’s tons of FX systems, but there are systems where you can do this one trade, where you’re getting both transactions and getting one price. I think there’s a ton to do on the automation side. How we do it, I’m not sure. But there really is an amazing amount of automation that can happen and a huge amount of savings,” he said.

Editing by Lukas Becker