Technological advances are shaping the way asset management firms operate, as they look for ways to introduce artificial intelligence applications to monetise data, and improve automation from the front to the back office.
Back in 2016, SEI wrote a white paper entitled The Upside of Disruption: Why the Future of Asset Management Depends on Innovation, in which it highlighted five trends shaping innovation: Watsonisation, Googlisation, Amazonisation, Uberisation and Twitterisation.
Witnessing the exponential changes occurring within and outside of the asset management industry as it relates to artificial intelligence, data management, platforms, social media and the like, SEI, in collaboration with ANZU Research, has updated these themes in its new series, The Exponential Pull of Innovation: asset management and the upside of disruption.
With regards to the first trend, Watsonisation, a lot has changed in terms of the power, sophistication and scale of artificial intelligence applications being used within asset management.
As the first of 5 papers in this series being released over the coming months, SEI’s new Watsonisation 2.0 white paper points out, “successfully harnessing technology in a complex and heavily regulated industry like ours is not easy. With new technologies and business models making change a constant, the financial services industry is being reorganized, re-engineered and reinvented before our eyes.” There are now dedicated AI hedge fund managers such as Aidiyia Holdings, Cerebellum Capital and Numerai, all of whom are pushing the envelope when it comes to harnessing the power of AI in their trading models.
According to a report by Cerulli, AI-driven hedge funds produced cumulative returns of 34 per cent over a three-year period from 2016 to 2019, compared to 12 per cent for the global hedge fund industry. Moreover, Cerulli’s research shows that European AI-led active equity funds grew at a faster rate than other active equity funds from January to April this year.
That trend will likely continue as asset managers tap into the myriad possibilities afforded by AI. As SEI notes, portfolio management teams are tapping in to the predictive capabilities by working alongside quantitative specialists with the skills needed to train AI systems on large data sets.
Large managers such as Balyasny Asset Management are now actively embracing a quantamental strategy to mine alternative data sets and evolve their investment capabilities. To do this, they are hiring ‘sector analysts’; people with sector expertise and superior programming skills in programming languages such as Python. The aim of this is to act as a conduit between Balyasny’s quantitative and fundamental analysts.
SEI argues that asset management is perfectly suited for the widespread adoption of AI.
They write: “Data is its lifeblood, and there is an abundance of historic and real time data from a huge variety of sources (both public and private/internal). Traditional sources of structured data are always useful but ripe for more automated analytics.”
Julien Messias is the co-founder of Quantology Capital Management, a Paris-based asset management that focuses on behavioural analysis, using systematic processes and quantitative tools to generate alpha for the strategy. The aim is to apply a scientific methodology based on collective intelligence.
“Our only conviction is with the processes we’ve created rather than any personal beliefs on how we think the markets will perform. Although it is not possible to be 100 per cent systematic, we aim to be as systematic as possible, in respect to how we run the investment strategy,” says Messias.
Messias says the predictive capabilities of AI have been evolving over the last decade “but we have really noticed an acceleration over the last three or four years. It’s not as straightforward as the report would (seem to) suggest, though. At least 50 per cent of the time is spent by analysts cleansing data. If you want to avoid the ‘Garbage In Garbage Out’ scenario, you have to look carefully at the quality of data being used, no matter how sophisticated the AI is.
“It’s not the most interesting job for a quant manager but it is definitely the most important one.”
One of the hurdles to overcome in asset management, particularly large blue chip names with decades of investment pedigree, is the inherent conservatism that comes with capital preservation. Large institutions may be seduced by the transformative properties of AI technology but trying to convince the CFO or executive board that more should be done to embrace new technology can be a hard sell. And as SEI rightly points out, ‘any information advantage gained can quickly evaporate, particularly in an environment populated by a growing number of AIs’.
“We notice an increase in the use of alternative data, to generate sentiment signals,” says Messias, “but if you look at the performance of some hedge funds that claim to be fully AI, or who have incorporated AI into their investment models, it is not convincing. I have heard some large quant managers have had a tough year in 2020.
“The whole concept of AI in investment management has become very popular today and become a marketing tool for some managers. Some managers don’t fully understand how to use AI, however, they just claim to use it to sell their fund and make it sound attractive to investors.
“When it comes to applying AI, it is compulsory for us to understand exactly how each algorithm works.”
This raises an interesting point in respect to future innovation in asset management. For fund managers to put their best foot forward, they will need to develop their own proprietary tools and processes to optimise the use of AI. And in so doing, avoid the risk of jumping on the bandwagon and lacking credibility; investors take note. If the manager claims to be running AI tools, get them to explain exactly how and why they work.
Messias explains that at Quantology they create their own databases and that the aim is to make the investment strategy as autonomous as possible.
“Every day we run an automatic batch process. We flash the market, during which all of the algorithms run in order to gather data, which we store in our proprietary system. One example of the data sets we collect is earnings transcripts, when company management teams release guidance etc.
“For the last four years, we’ve been collecting these transcripts and have built a deep database of rich textual data. Our algorithms apply various NLP techniques to elicit an understanding of the transcript data, based on key words,” says Messias.
He points out, however, that training algorithms to analyse textual data is not as easy as analyzing quantitative data.
“As of today, the algorithms that are dedicated to that task are not efficient enough for us to exploit the data. In two or three years’ time, however, we think there will be a lot of improvements and the value will not be placed on the algorithms, per se, but on the data,” he suggests.
Investment research is a key area of AI application for asset managers to consider, as they seek to evolve over the coming years. Human beings are dogged by multiple behavioural biases that cloud our judgment and often lead to confirmation bias, especially when developing an investment thesis; it’s the classic case of looking for data to fit the theory, rather than acknowledging when the theory is wrong.
AI systems suffer no such foibles. They are, as SEI’s white paper explains, “better able to illuminate variables, probabilistically predict outcomes and suggest a sensible course of action”.
Messias explains that at Quantology they run numerous trading algorithms that seek to exploit investment opportunities based on two primary pillars; one is behavioural biases which exist in the market. “We think our algorithms can detect these biases better than a human being can,” states Messias.
The second pillar is collective intelligence; that is, the collective wisdom of the crowd.
“We have no idea where the market will go – this is not our job,” asserts Messias.”Our job is to deliver alpha. The way markets react is always the right way. The market is the best example of collective intelligence – that’s what our algorithms seek to better understand and translate into trading signals.”
One of the truly exciting aspects to fund management over the next few years will be to see how AI systems evolve, as their machine learning capabilities enable them to become even smarter at detecting micro patterns in the markets.
Google’s AlphaGo became the first computer program to defeat a professional Go player without handicaps in 2015 and went on to defeat the number-one ranked player in the world. As SEI observes: “Analysts of AlphaGo’s play, for example, noted that it played with a unique style that set it apart from human players, taking a relatively conservative approach punctuated with odd moves. This underscores the real power of AI. It is not just faster and more accurate. It is inclined to do things differently.”
Logic would suggest that such novel, innovative moves (ie trades) could also become a more prominent feature of systematic fund management. Indeed, it is already happening.
Messias refers to Quantology’s algorithms building a strong signal for Tesla when the stock rallied in September last year when the company released its earnings report.
“The model sent us a signal that a human being would not have created based on a traditional fundamental way of thinking,” he says.
Will we see more hedge funds launching with AI acting as the portfolio manager?
“I think that is the way investment management will eventually evolve. Newer firms are likely to test innovations and techniques and if AI shows they can become more competitive than human-based trading, then yes I think the future of investment will be more technology orientated,” concludes Messias.