There are a number of perspectives on how data science, artificial intelligence and machine learning could be used to improve client outcomes and experiences.
One can look at the actual application of machine learning and data analysis in order to run money; for instance, hedge funds are increasingly using algorithmic trading strategies in a bid to enhance investor returns.
You might also look at the growing use of robotic process automation (RPA) and some AI and ML applications to improve the operational efficiencies of how financial services practitioners – including advisers – go about their business. This might be where so-called robo-advice would sit, for instance. Another view is based on client experience.
Let’s take these one at a time.
AI-based trading strategies are not exactly rare, nor brand new. According to a survey of hedge fund managers, more than half use AI and ML to inform their investment decisions, two-thirds use AI and ML to generate trading ideas. Just over a quarter use automation to execute trades and a third have been doing so for nearly a decade.
These are not necessarily new concepts but have certainly gained traction as our thirst for – and, through the internet of things, our ability to gather – data on a constant basis has grown so exponentially. According to BNY Mellon, more data can now be captured in one day now than in the entire 1990s.
As such, AI systems can tap into satellite imagery, global capital flows, point of sale systems and social media feeds. Cloud computing is pushing down costs and hedge funds have tended to lead the investment world in product development, ranging from new ‘pure play’ AI-based offerings, to established players hoping to enhance their traditional strategies with data-science-led techniques.
Like many (non-ML) quantitative or rules-based models, these funds seek to find inefficiencies in historical markets, leveraging them to find small slivers of return on an ongoing, reliable basis. The underlying approaches might be similar, but the mathematics used to derive each type of model will differ.
However, for the typical retail investor, I hold some reserve over suitability. Their inherent opacity, difficulty in articulating exactly how they are investing and the challenge of presenting an accurate assessment of where the risks lie versus the potential return prospects do not sit well with me – or the regulator, for that matter.
I just cannot see us getting to a point where a purely technical, algorithm-based, ML-driven product would be easily explainable to a retail investor. Certainly not in the short to medium term. Maybe it never will. Or perhaps I am just a bit of a stick-in-the-mud.
Where I see more viable opportunities is where data analysis can be used to help advisers create plans and approaches for clients.