Five years in the making

This year, we at UOB Asset Management (UOBAM) will soon be celebrating the fifth anniversary of our AI-Augmentation journey. If, as some are saying, AI is opening a “new frontier in investment management”, then we can be described as one of its founding pioneers in Asia.

Indeed it appears that there is growing interest in the potential of AI tools to generate outsized returns. However to date, there are few other asset managers with a track record as long as ours, and therefore few with learnings and experiences as rich as ours.

So this is an opportunity for us to address some of the myths and assumptions that have emerged about what AI can (and cannot) do for investors.

Lesson #1: Yes, AI really can deliver value

There are several US academic studies that have researched whether AI can beat the market, and the results appear to be mixed. However, having tracked the performance of our AI-Augmented funds for five years, we would say that the case for AI is a convincing one.

Our experience is that our AI-Augmented techniques work particularly well in markets like Asia where there are a very large number of under-researched stocks, and very diverse macroeconomic conditions. Two of our Asian equity funds apply our AI-Augmented techniques in full and they currently carry Morningstar’s highest 5-star rating. In 2023, both funds beat their respective benchmarks by 6 – 8 percent.

Lesson # 2: Paradoxically, humans are key to AI

When AI-driven investing is being discussed, many investors fear that artificial intelligence will completely replace human intelligence and investment management decisions will be made entirely by machines.  Our experience is that it is the effective balance between human and artificial intelligence that has generated the most alpha for our funds.

This is because, on the one hand, AI-driven stock selections help to overcome human biases that can cause certain companies, sectors or markets to be overlooked. On the other hand, humans are crucial for preventing undesirable holdings, and avoiding AI becoming a “black box”. We ensure all AI-based stock picks are verified by human analysts before including them into our portfolios. 

Lesson #3: AI doesn’t mean invincible

We note that it is unrealistic to expect AI-Augmented techniques to deliver strong positive performances every single month. There are of course months when our AI-Augmented funds do underperform their respective benchmarks. However, we observe that these funds tend to outperform the markets more frequently than underperform.  Furthermore, on average, the size of the outperformance tends to be larger than the size of the underperformance. 

This is because, to help ensure maximum alpha, our model is designed to seek out, not just outperformance, but where the greatest opportunities lie. These stock picks are made independent of any asset allocation considerations and account for over 90 percent of the funds’ outperformance.  

Lesson #4: AI-augmented funds are less, not more, volatile

It is often assumed that AI-Augmented funds, given that they are capable of delivering higher returns, do so at higher risk. We have not found this to be the case. Unlike traditional linear methods of predicting returns, our AI-based, non-linear predictions do not pursue a specific hypothesis. Instead, we use a pattern recognition technique which means our models are less likely to be side-tracked by one-off events and immaterial market noise.

This helps to improve our accuracy when predicting stock corrections and rallies. As a result, our AI-Augmented funds tend to show lower volatility than their respective benchmarks or their peer group averages.

Lesson #5: Technology is only part of the story

It is often said that active investment management using AI is a self-defeating exercise. Like most technologies, AI models will eventually be replicated by other professional investors and become widespread. Therefore, the returns advantage that they provide is merely temporary.

In our view, the technology and AI models underpinning our AI-Augmentation strategy is less important than the investment process that surrounds them. There are hundreds of open-source AI models available for free and many more available for a price. And yet, the ability of these models to successfully outperform the market is still relatively limited and far from universal.

In our case, factors such as management support, a culture of innovation, organisational structures and constant process refinements over a prolonged period has got us to where we are today. As with all valuable activity, practice makes perfect.