In which areas could AI be most useful in the FX markets?
Bart: People sometimes try to bring AI into areas where there is limited value or to showcase its use rather than to bring value to the front end. Where I think it’s really helpful is where there’s a vast amount of information you need to absorb and then base your decisions on a deducted answer. AI can help to find those correlations. I would suspect it’s still a little too early to create any real trading models or decision making as the regulatory framework on controls and adoption of AI is undefined. However, reporting, TCA, idea generation, that’s where I see AI coming in, where the machines can look for correlations over a much bigger piece of data. It is often underestimated how useful AI is in support.
In the FX market, formerly you would need to get to credit applications post-trade to reduce the risk, which was a major exercise via manual processes. This is where the AI evolution is likely to accelerate because the technology lends itself to it. This is where optimisation comes in, but you could also think about sending out those trades automatically. The amount of data that you can progress within an AI model is much greater than at a human level or structurally.
What’s the next stage of automation in FX?
Neill Penney
Neill: We’ve come off a few years of very high volatility and high trading profits. The markets have quietened this year, most of our customers are now looking at how they can do things more efficiently and at lower cost, and that is going to be the main driver for increased automation. At the same time, risk tolerance is going down and the requirement that people execute with provable methodologies – where they can establish post-trade that they executed in accordance with their company’s policies – has gone up.
With the next iteration of trade execution, customers will need more automation, but it will also need to be backed by quantitative analysis and reasoning. Customers want advice when they execute and, ideally, they’d like that advice to translate into setting up the order in the system. What they need is a model that will say “having analysed multiple years of trade history, we recommend you use this algo, or this execution strategy, to execute a particular trade.” It’s about taking all that data and historical insight, and then turning the analytics into a tool that guides the trader towards the best way to trade. In a year, AI has gone from a buzzword into something that people are making use of in their day jobs.
Bart Joris
Bart: What you will also see – and you can see this with NDFs – is less liquid products are going to become more transparent, and the more transparent they are the more trading you get. The more trading you have, the more automation you can drive into it. I would expect the biggest evolution to be in the Forwards market. It could also be FX options or swaps where there is still manual trading, that’s where you will see acceleration of technology.
In swaps, the dealer to client market is already electronic, though the interbank market is still very manual. There are several reasons for this as impact on capital and position skew is inherently part of the management of the forward books. As more data becomes available in a market neutral curve, it becomes much more accessible for everybody to trade into. That will also clear a lot of the risk. For options, it’s probably a little bit further away, but in the forwards market, there could be a lot more automation in the next two or three years.
Neill: A year ago, we added an API onto our forwards matching venue to allow market makers to trade electronically. Following the rollout, we had further feedback from customers that to take maximum advantage of the API technology they needed some other capabilities. For example, electronic control for credit management. To this point, currently, in the interbank forwards market people agree a price and then they manually check whether they have sufficient credit. That’s fine as a manual trading convention but it doesn’t work for automated trading. In the next generation of forwards matching, which we’re launching next year, this automated credit checking – so called “hard credit” – is built in. As the market sees the opportunity here, we’ll start to see behavioural changes in people’s trading once all the technology pieces have been delivered.
LSEG confirmed plans to launch a new FX matching platform and announced a new artificial intelligence-based dealer tool in its H1 2023 results. Why now?
Neill: We’ve been rebuilding and reinventing our FX product set since 2018. The process began when our business separated from Thomson Reuters and became Refinitiv. Our first step was working on the transition from Eikon into Workspace. Subsequently, becoming part of LSEG gave us the opportunity to invest in our foreign exchange venues. We are now in a significant multi-year initiative to upgrade our venue technology with LSEG’s exchange technology.
The first thing we’re bringing to market as part of this venue transformation work is our NDF Matching venue. It showcases a several of improvements to our FX capabilities. It’s the first time we have supported NDFs in Matching. It’s our first venue in Asia. It’s got clearing built into the execution mechanics. It’s the first time we’re bringing together our FX post-trade expertise from LCH ForexClear and our execution expertise from Matching and FXall. And finally, we’re launching a new front end for it, built into our new Workspace desktop.
Why NDFs first?
Neill: NDFs are a growing part of the FX market, with limited customer options when it comes to execution on an order book. As such, there has been clear interest from customers for us to support NDFs as part of Matching. There is also a lot of interest from customers in the cleared execution part of the venue which should result in improved liquidity, more efficient use of credit, and reduced administrative overheads.
In the wider context of our technology transformation, another benefit is the ability to deliver and rollout the new technology for a smaller venue than our core spot and forwards Matching platforms. Our customers have been very supportive of this phased approach, which lowers the risk and complexity of the technology transition for all players.