Banks, vendors mine AI for corporate FX hedging

The inexorable advance of artificial intelligence across banking’s backyard continues. Now, banks and technology vendors are developing machine learning tools to help support non-financial corporates with their foreign exchange hedging decisions.

For corporate treasurers, most of their FX hedging is carried out using manual processes. Many companies do not have the same advanced technological processes, data scientists and trading strategies that the more sophisticated participants in the FX markets possess, meaning they are more exposed to human error.

With currency markets feeling the effects of greater volatility, corporates are looking to improve the way that they hedge their FX exposure.

In a recent survey by U.S. Bank, 39% of chief financial officers said they are not confident about their ability to manage and mitigate new risks. Over half of respondents also said they needed to improve their FX hedging strategies.

The response has been to turn to AI and machine learning tools that can support their FX hedging needs. The U.S. Bank survey showed that 51% said they are prioritizing investing in AI in the finance function, with risk management as their primary motivation.

Some FX dealers have ventured into this new technological realm to improve the way corporate clients handle their data, and are now starting to look at the next iteration of this tech. “Today we’re using AI for data validation and trade review. In addition to enhancing these pre-trade applications, we’re also considering ways to incorporate AI in relation to the execution process,” says Steven Fenty, global head of currency management at State Street Global Markets.

There’s still a fair amount of healthy skepticism on using generative AI to make trade recommendations or using machine learning to execute trades on behalf of a client

Nick Pedersen, NatWest Markets

Over the past few years, treasury teams at many large companies have started experimenting with algorithms that aim to reduce the manual labor involved in some of their more complex, time-consuming tasks, including cashflow and exposure forecasting, and the identification and execution of risk management trades.

They are now hoping to leverage the technology’s ability to analyze vast amounts of structured and unstructured data, including historical price movements, market sentiment, and macroeconomic indicators, to accurately assess currency risk exposure and predict future market trends.

“We are starting to see corporates looking at implementing those [AI] tools as a way of getting better data access centrally, in a faster and cheaper manner than spending a long time building or implementing a treasury management system across their various entities,” says Nick Pedersen, global head of digital at NatWest Markets.

The UK bank has also begun trialing this technology for creating trade recommendations, trade summaries, portfolio summaries and using generative AI in chats.

“We apply machine learning tools to analyze client portfolios. And that’s becoming quite interesting, when you do simulation scenario analysis and put those tools into the hands of a customer themselves, rather than just replaying the results back to them,” says Pedersen.

Vendors are also developing AI-based services to help corporates with their hedging choices. One example is a new platform from C8 Technologies, led by two former BlueCrest Capital Management partners. The C8 Hedge platform uses machine learning and statistical models to predict future FX movements.

The platform’s predictive model collates macro data such as purchasing power parity, as well as movements in FX spot, commodities, fixed income, oil, and precious metals markets to build expectations on currency relationships. The model then uses these outputs to calculate optimal FX hedge ratios for corporates’ currency exposures and add risk weights to each currency exposure.

“We aim to identify relationships between this data and action in the foreign exchange market. Some of these relationships are strong, while others are weak, therefore, our task is to use regression and expectation engines to determine which relationships are the strongest,” explains Ebrahim Kasenally, head of research and co-founder of C8 Technologies.

By comparing predicted currency movements with reality, the tech firm is able to evaluate the strength of a trend for each currency. The model then translates these trends and forecasts into an actionable hedge ratio for each asset and liability exposure that the client holds in a foreign currency.

Research by the company shows that the optimal timespan for corporates to review hedging activity is one month: that’s the “sweet spot”, says John Webb, head of C8 Hedge.

Balancing act

One useful technique for corporates is to dynamically adjust hedge ratios, enabling firms to generate returns when currencies appreciate. However, many corporate treasurers—particularly at smaller firms—have neither the experience nor resources to achieve this.

Hedge fund giant Millennium Global Investments offers an outsourcing service which aims to optimize a client’s currency exposures by dynamically increasing hedging in periods of foreign currency weakness and reducing hedging in periods of foreign currency strength.

The service uses AI and machine learning algorithms to help rebalance the hedging ratios.

“It combines both macroeconomic underlying inputs as well as market dynamics, price changes, and alternative data sources to generate signals, which are then used to automatically vary hedge ratios based on time constraints,” says Eric Huttman, CEO of MillTech FX, the fintech arm of Millennium Global.

The firm is also developing a currency hedging tool, where the machine learning technology identifies risk signals on a particular currency pair and then feeds that information into a corporate client’s broader decision-making process.

Depends on the data

However, experts caution that machine learning-based models and regression techniques are not a magic bullet for FX hedging. As with artificial intelligence in general, the technology is limited by the quality of the data inputs. Or, as computer aficionados like to say, garbage in garbage out.

“If you can’t go back historically for a long time series, or you don’t have as many data points to input, then machine learning becomes not as effective and a little redundant,” says NatWest’s Pedersen.

Garth Appelt, head of derivatives, FX and emerging markets macro trading at Mizuho Americas, is skeptical whether AI-assisted hedging will be beneficial to clients in all instances—at least, for now. Longer-term hedges would require corporates to produce accurate forecasts of, say, future cashflows versus earnings.

“If you don’t know what your earnings are, then you can’t do long-term hedges. So, you’re not just predicting the market, you’re predicting your own business,” he says.

Corporates also might be wary about divulging sensitive information regarding their firm’s financial performance to service providers.

Possibilities of hands-free and automatic hedging using AI are still some way off, then. But what is more realistic is using the technology to simplify and automate the elements that go into making those hedging decisions.

“There’s still a fair amount of healthy skepticism on using generative AI to make trade recommendations or using machine learning to execute trades on behalf of a client,” says Pedersen. “I think the value in new waves of AI is going to be about data collection, rather than decision making, and it’s going to be about all of the tasks associated with coming to a decision or setting a hedging policy.”

Looking ahead, State Street’s Fenty believes the next phase of the technology will relate to how it can shape execution choices, such as forecasting the best time to execute, adapting forward tenors, and recommending the best type of hedging instrument to use.

“These are decisions that have to be made daily, and AI would help make those decisions more quickly with a broader dataset.”