HSBC Holdings has put artificial intelligence to a novel use: It’s using different forms of AI to analyze data, choose assets in which to invest, and rebalance the asset allocation weekly for an index called AiMAX, which it launched last week.
AiMAX is geared toward clients of HSBC’s wealth businesses where it can be accessed through HSBC structured products as well as insurance and fund vehicles offered by third-party distribution partners. The underlying technology, a combination of the AI and discovery built into IBM Watson and AI modeling software developed by EquBot, can analyze a wide range of data sets, including satellite images of farm crops, shipping containers in the ocean and foot traffic patterns at shopping malls, alongside more traditional data sources like economic reports, news feeds and social media posts to make investment decisions.
HSBC hopes its AI-driven products will deliver better performance to clients.
“We're trying to offer better indexes with better investment performance, which helps investors grow wealth and save for retirement,” said Dave Odenath, head of quantitative investment solutions for Americas at HSBC, which is based in London and has $2.918 billion of assets. “AI is doing the job a team of analysts would have historically had to do. Being able to leverage that technology and plug it into some of our strategies allows us to do more with less.”
It's the latest example of a growing trend in which financial services firm deploy AI to do or to enhance work normally done by humans.
“What's great about this technology is it has the ability to crawl anything and everything in the public domain,” Odenath said. “So things like social media that potentially have had an impact on the markets are covered by the AI. The system's ability to read social media gives it a bit of alpha on the sentiment.”
AiMAX, its creators say, can simulate the work of a large team of global market analysts and traders to come up with market insights that can be used in creating and rebalancing investment portfolios. Principles of modern portfolio theory are combined with predictions generated by AI. The AI software rebalances portfolios every week by following a three-step process. First it provides forecasts based on data. Then AiMAX tests each possible combination of the 15 investable asset classes. Then it selects the portfolio likely to generate the highest return.
“The AI forecasts determine the asset allocation for AiMAX and the individual equities for AiPEX going into the index portfolios,” Odenath said.
Some hedge funds, including Renaissance Technologies in East Setauket, N.Y., which has recruited IBM Watson engineers, use AI in their investment decisions. HSBC is the first company to create AI-driven indices.
HSBC first did this with an AI-driven stock market index called AiPEX it launched in August 2019. According to HSBC, the AiPEX Total Return Index outperformed the S&P 500 Total Return Index by 7.60% in 2020 and has outperformed by 4.14% since launching.
AiPEX has an investable universe that's equivalent to the Russell 1000, an index made of the 1,000 largest U.S. companies by market cap. On a monthly basis, the AI is asked to pick a portfolio that's typically around 250 stocks that it predicts will grow the most over the coming month.
AiMAX is a little bit different because it has a defined investment universe of 15 different asset classes that it can invest in. On a weekly basis, AI-generated price forecasts are used to come up with a portfolio expected to appreciate over the coming month. A portion of the portfolio is rebalanced each week.
The work is similar to what a financial adviser would do. The index is implementing an asset allocation model, trying to provide growth for investors while controlling for risk, and trying to rebalance for different market conditions.
“I like to think of AiMAX as taking some of the best parts of a modern portfolio theory approach to asset allocation and trying to enhance it,” Odenath said. “We're not trying to rewrite the playbook on asset allocation. We're just trying to augment it with this AI technology.”
But where humans might use what’s happened in the past to rebalance a portfolio, the index is using AI to forecast the future performance of each of the underlying assets. Those forecasts dictate what the asset allocation looks like.
The AI analyzes millions of publicly available data points each day.
“It could really be anything from macro data like employment rates and inflation targets to things like social media posts about a particular company or product,” Odenath said. “The power of the technology and why we think it's so useful is not only can it keep up with the amount of data that's being generated each day, which is growing on an exponential basis, but then also make sense of it. You have all these different data points that are coming in. Reconciling that, making the connections, finding the patterns between those is extremely challenging. I think it's almost beyond human comprehension.”
It could take thousands of human analysts to do the work being done by the AI system, he said.
“We like to say that each one of those thousand stocks is essentially being followed by a virtual stock analyst that specializes in that particular company,” Odenath said. But the AI can also see connections, patterns and themes among the stocks that individual human analysts would struggle to identify.
Athene Americas offers an annuities index linked to AiPEX and AiMAX. HSBC also offers the AiPEX and AiMAX indices in structured products that are being sold by financial advisers.
The role of IBM Watson
IBM’s Watson technology contributes to the data going into the AI.
IBM Watson Discovery can comb through the language in vast amounts of news articles, social media posts and documents in multiple formats (including PDFs, charts, tables, images and graphs) to pinpoint precise answers and identify trends and patterns in content.
“Having good clean data, which IBM has, is extremely important when you're trying to train an AI over time,” Odenath said. “Watson is helping us find patterns in the connections between different types of data. Where they're really specialized is in nontraditional types of data, things that aren't going to appear on spreadsheets. Things that aren't just numbers that you can plug into a model.”
Watson's databases of historical information were also used to back-test the index for the last 15 years to set expectations with investors as to how it would perform.
EquBot's role
EquBot spent several years in an IBM incubator figuring out how to turn IBM’s Watson technology into a stock picker and asset allocator. HSBC asked the San Francisco company to create the new indices, which are rules-based.
EquBot's platform can analyze more than 50,000 global companies as well as commodities, government interest rates and market signals. Market prices, 10-K and 10-Q filings, industry reports and tweets can all become part of EquBot’s knowledge graph.
“We're looking at all of these different companies, these millions of different news articles in dozens of different languages each day, to develop different scores for the financial health of a company,” said Chris Natividad, chief investment officer at EquBot. Management scores help determine if different companies or economies have had a strong performance record during different market environments.
In addition to the data it receives from IBM Watson, EquBot uses data from industry reports and financial data providers.
“Watson allows us to more effectively process news and discover patterns within the different forms of data,” Natividad said. “There are elements within IBM Watson that also allow us to ensure that there's no information bias, to detect different forms of drift in AI models, or if too much data becomes overlapping.” In model drift, the relationship between the target variable and the independent variables changes with time, which makes the model unstable and its predictions faulty.
Natividad sees an AI arms race going on among financial firms.
“A couple of decades back, we used to think about quants on the trading desk," he said. "A lot of these financial firms are going through this: Do we build it, or do we buy this type of asset to bolster our operations?”
EquBot has been meeting with endowments, sovereign wealth funds, registered investment advisers and individual retail investors interested in using AI in investment management operations.
“It's our belief that by 2040, about 90% of the investment management applications out there will use AI in some form or another,” Natividad said.
