Ai for alpha: issuers are using machine learning solutions to develop new products

Not surprisingly, a number of AI companies have turned to machine learning to develop asset allocation and risk modelling strategies as well as to unlocking alpha in large datasets.

Ai For Alpha entered the structured products market in 2022 on the back of a partnership with Société Générale to develop a new underlying for the US indexed annuities market – the French  financial software company was also appointed recently by asset management company Artificial Intelligence Asset Management to provide asset allocation advice for global macro funds.

SRP spoke to Beatrice Guez (pictured), CEO of AI for Alpha, about the use of artificial intelligence and machine learning to develop new dynamic allocation strategies for structured products.

Our clients are using the Ai 360 Portfolios’ signals to implement their own risk-on and risk-off indicators on equity markets - Beatrice Guez, AI for Alpha

Ai for Alpha is using its Ai 360 Portfolios solution to develop different strategies for major French and US investment banks, including underlyings for the structured products market.

One of the strategies used the company’s 'decoding machine learning technology’ to build a basket with the best trend follower CTAs in the market.

“The technology allows to replicate a fund or a strategy based on its NAV, using a list of predefined factors,” said Guez. “The aim is to provide a more liquid, transparent, and cost-efficient solution than investing in traditional CTA funds.”

Taking a closer look at the strategy, the model allows to decode CTA’s performance with a very high correlation – 80 to 90%.

“[This] is very interesting for investors to have insight on their different positions across asset classes and markets (equity, bonds and FX),” said Guez. “Investors can understand how CTAs made most of their performance last year at different points in time.”

The Ai for Alpha model shows that in 2022, for example, most of CTA’s performance came from short-fixed income and long dollar positions.

“We used this technology to decode major global balance European funds, risk parity funds and global macro funds,” she said. “Clients can either invest in the strategy via an index-linked investment or just use the information as research or qualitative information.”

Other indices developed by Ai for Alpha for issuers of structured products are based on market regime anticipation across different asset classes, especially equity but also commodities and fixed income.

“Our clients are using the Ai 360 Portfolios’ signals to implement their own risk-on and risk-off indicators on equity markets,” said Guez, adding that based on this regime prediction, the model was long equity in 2021, and started to short at the end of 2021.

“Last year was mainly bearish until November 2022 when the model started to take a long position again,” she said. “This is a very interesting approach for investors looking for risk-on and risk-off indicators as it's different from a quantitative approach and it's based on rich supervised learning.”

What is advantage of using a machine learning approach vs. a quantitative approach?

Beatrice Guez: The AI approach leverages a large amount of market information such as technical indicators, financial metrics, micro information and macroeconomic data. This information and the different types of signals are difficult to digest by humans. Our technology uses machine learning to process historical data to link the financial information with the performance of different asset classes.

This approach can be used to either develop quantitative strategies and or for qualitative information.

For each market, our machine learning model analyses the evolution of a lot of different features (interest rates, equities-bonds correlation, economic surprises…) to provide the probability of a short-term crash (1month prediction). We can see red flags (bear features), and green flags (bull features) at each point of time. Besides the market regime prediction on each asset class, the model provides a global allocation taking into account the market conditions. 

For example, the model was rather bull last March despite the SVB and banking turmoil, with a long position on Equity market. This means that the risk of a short-term crisis in the equity market was rather low. Clients were able to understand which features that were driving this result.

What opportunities and applications do you expect for AI/ML as they become more visible in the financial world?

Beatrice Guez: AI can help our clients decrypt financial markets. We see a very strong interest in decoding thematic investments. In the US and Asia, people are very interested in CTAs. In Europe, people are more interested in Global Balanced or Risk Parity funds, because this is where they are usually investing.

We're not an asset manager, we work with financial institutions such as investment banks that are using the model to develop new investment solutions for their clients including structured products.

More retail investors are looking for quantitative strategies and the strategies we develop meet that demand.

For example, the global balance fund decodes a basket of 50 largest European funds. These are the names that ring a bell with European investors.

Can AI and ML help to develop new ESG compliant strategies and prevent greenwashing?

Beatrice Guez: On ESG, the decoding technology is applied on the universe of ESG funds. We can replicate a fund based on macro factors, in which markets are invested as a main exposure, growth/value tilts on equity and ESG factors. In this case, we help our clients understand the main drivers of funds in terms of ESG investment and to classify a fund by ESG intensity. The model shows the fund with the highest ESG score and intensity level and we can understand the main factors driving this ESG investment. We can help our clients have a better understanding of how each fund is positioned in terms of ESG investment.

In order to run the analysis, we need to have daily NAV with at least one or two years of data. This technology is applicable for any type of strategy, including structured products, as long as we have daily data.

Are there any risks in using AI to develop new financial products? What are the pros and cons of using AI in finance?

Beatrice Guez: At this stage we still need humans to make the final decision. The model provides signs to understand the factors driving a strategy, but the manager decides. This technology is a complement as opposed to a replacement and we need to be reactive to unexpected events like a black swan.

AI is quicker and allows you to leverage information at different levels.  The advantage is that AI not only allows you to analyse large amount of data and provide market signals, but also to take client constraints into account and assess different type of scenarios in the portfolio construction stage.