In the past decade, the asset management industry has witnessed a significant increase in the use of artificial intelligence (AI) and machine learning. Asset managers now have access to a large amount of information that can used in capital allocation decisions. This raises a number of questions. First, do asset managers have access to a reliable sources of information that allows them to allocate capital in a way that would generate positive risk-adjusted returns? Second, do funds relying on AI-enhanced strategies over perform?

In a first paper presented by Roméo Tédongap (ESSEC Business School), the authors relies on  a Stock Active Share (SAS) measure and analyze the risk-return characteristics of portfolios formed by ranking stocks according to their active shares among mutual funds. Then, they design a “SAS-Oracle strategy” by using four different machine learning models to anticipate the stock’s SAS value in advance. This allows the authors to construct investment portfolios proactively. They show the efficacy of machine learning-based investment strategy and that investors can enhance their portfolios by learning from mutual fund managers’ historical holdings, which reflects skilled decision-making. 

The second paper presented by Mauricio Praxmarer (University of Innsbruck) focuses on investment funds that apply certain AI-enhanced strategies. The authors ask whether i) funds can adopt AI-enhanced strategies, ii) if so, do they provide investors with any advantage and iii) if not, is the AI tag used for “AI washing”? After analyzing the performance of AI funds, the authors investigate whether the observed return is due to skill or luck. This is done by decomposing a fund’s excess return into a skill component capturing the skill level of the fund and an activity component which captures how actively a fund exploits the set of skills. Finally, the authors investigate how money flows in and out of AI funds.