By Paul Stafford
For any business with interests overseas or funds with foreign investors, managing Foreign Exchange (FX) risk should be a vital part of the overall financial strategy, with swings in currency able to make or break a cross-border investment. However, managing an FX hedging program has traditionally been difficult and time-consuming for most and even when strategies were applied, the results did not always meet expectations.
In recent years, this has started to change. Now technology aids the selection and application of traditional hedging solutions such as derivatives or allows businesses to run vast simulations to build a complete picture by evaluating the performance of every strategy against every possible scenario.
However, this is just the beginning. Machine Learning (ML) and, in particular, Artificial Intelligence (AI) are currently getting a lot of coverage for the potential benefits and risks they can bring to every element of human life. Yet what has received relatively little attention is how these innovations can and will be used to aid FX hedging. At Deaglo, we’re already utilizing Machine Learning (ML) technologies to build innovative hedging solutions for funds and businesses, no matter the industry, location or currency. But there’s so much more to come – especially once AI’s hedging potential is realized.
In this blog, we’ll examine how the two technologies are currently being used in FX and explore the future of FX hedging.
What we know now
AI’s time is now (and forever)
AI is not a new concept and has been around in one form or another since the 1950s. However, in recent years it has become an increasingly ubiquitous part of our lives, even if many of us didn’t realize it. However, that all changed last December with the launch of OpenAI's ChatGPT3. As Jeremy Kahn wrote in Fortune: "A few times in a generation, a product comes along that catapults a technology from the fluorescent gloom of engineering department basements, the fetid teenage bedrooms of nerds, and the lonely man caves of hobbyists—into something that your great-aunt Edna knows how to use."
The Large Language Model’s (LLM) release was the headline act in a generative AI explosion that also saw the launch of OpenAI's image-producing DALL-E 2, Midjourney, Google and Meta's video production AIs, Baidu’s Ernie Bot and Google’s Bard in just a few months. Since then, there has been somewhat of a gold rush with funding to AI startups bucking wider investment trends to receive billions of dollars in venture capital.
Despite this, in the FX and FX hedging industries, the application of AI is still relatively underdeveloped. While some in the industry claim to be offering it to aid hedging or trading, many are incorrectly – or disingenuously– using it as a synonym for automation or ML. Even so-called ‘algorithmic robo-advisers’ are mostly just client onboarding tools that score customer risk preferences and before connecting them to a predefined segment offering. To date, the most exciting hedging technology exists in the ML space.
ML FX is already here
While ML is overall more limited in its scope for FX hedging, its self-learning algorithms can be used to great effect and the technology is already far more developed in the space. Natural Language Processing (NLP), for example,is used to analyze text data to extract key information and insights about currency markets. This can include parsing through news articles, central bank statements, and other sources to identify trends and signals. In a similar fashion, ML can be used to analyze news articles, social media feeds, and other sources to assess the sentiment of the market and identify potential risks or opportunities. It can also be used to analyze the risk associated with certain currency pairs and to predict market trends and price movements in real time.
At Deaglo, we’re using ML in three key ways:
Proxy identification
Some currencies are so volatile or have no derivatives market, they are unhedgeable. We use ML to figure out what other currencies track that unhedgeable currency and then hedge those as a proxy currency. This means the gain or loss in the hedging of the proxy currency approximates the gain or loss in the unhedgeable currency.
Contingent hedging
Often with cross-border purchases of certain assets, offers are contingent on due diligence. This can take months and within that period currencies can fluctuate, causing the value of purchase to shift. In that scenario, there are no derivatives that are practical to use. So through ML, we are able to create a basket of currencies that exhibit minimal variance between the buyer’s and seller’s currency.
The buyer then puts their money into a basket of optimally weighted different currencies and at the end of the due diligence period, if the deal goes through, they're converted to the seller's currency. If the deal falls through, they're converted back to the buyer's currency. The minimal variance means the buyer shouldn’t exhibit a loss over the due diligence period.
Over/under analysis
We utilize Long Short-Term Memory (LTSM) Neural Networks to help evaluate the macroeconomic pressures on a currency pair and determine whether a currency is over or under-valued. LSTMs are used to learn, process, and classify sequential data and by feeding it past macro data such as interest rates and Credit Default Swap rates (CDS) from the countries of the currencies involved, we're able to predict and compare the market rate with the model rate and assess its true value.
What’s to come
These are still early days for AI and ML in FX hedging. However, the functionality of both technologies is already so defined it’s possible to hypothesize future applications within the FX hedging space.
LLM administrative support
AI uses all kinds of structured and unstructured data to perform complex tasks. Generative, large language models like ChatGPT4 can be used to automate back-office administrative tasks such as the onboarding of clients, KYC and AML assessments, and policy document and report generation. It will also be able to confirm whether reporting and procedures are compliant. For managers and advisers, this will all allow them to scale quicker while also handing them back hours of time to focus on more important tasks.
LLMs also have the capability to identify patterns and trends in fundamental data and offer suggestions on how best to respond when hedges are deep in or out of the money.
24/7 white glove service
Hedging and FX, although heavily automated, still rely on a human element. At Deaglo, although we’re very much a technology-first company, we still talk to our clients to understand their needs and the situation. Clients appreciate this. However, this comes with constraints. Primarily, a lack of time. Applied AI can solve this. In the future, clients will be able to talk to the software and outline the situation and it will be able to present clients with the best options to hedge. It can be fed more and more information and will be monitoring thousands of factors 24/7, running simulations, changing and improving its predictions and analysis as it goes.
The AI will be able to understand 100s of languages, allowing it to interact with clients from all over the world but also analyze a global array of sources. Not just that, but it will be able to ascertain tone, uncertainty, hesitation and more to offer a far deeper strain of analysis.
Black swan protection
Black swan events can drive currencies crazy, driving what were far out-of-the-money short positions into the money, costing the hedger and driving derivative position valuation out of the money. Although a human currency advisor can advise on a position, if the event takes place out of working hours it could be too late to act. With an AI hedging solution, clients can give the AI a mandate to handle the risk and make decisions on their behalf in the case of extraordinary volatility.
The start of something special
We’ve laid out some solutions of what the near future could look like but above all, where we see the greatest benefits currently, are in the time that AI and ML can dedicate to hedging and the speed with which it can execute analysis and offer solutions. Current digitalized hedging solutions are remarkably effective, the main issue is lack of human time, brain power and speed. AI has no such restrictions. Working 24/7 it can run hundreds of thousands of simulations concurrently, figuring out the optimum solution at the optimum time.
ML and, in particular, AI are already changing the world, but this is just the beginning. Across the financial and business world, people are looking at how technologies can lower risk, increase efficiency and drive growth. Foreign Exchange is no different and a completely automated FX risk program offering operation efficiency and speed is achievable in the near future. At Deaglo, we’re already looking at how we can add AI to our services.
The most exciting thing, however, is that the most exciting future use cases have likely not even been thought of yet. And when they are thought of, it may not even be humans that figure them out, it will be the AI. As OpenAI CEO Sam Altman said when he was asked by investors how ChatGPT would eventually make money: “basically we will ask it to figure out a way to generate an investment return for you.”
Comments