Scientific Beta

An interview with Maxime Ricomes on the topics of machine learning and AI following his participation at the roundtable at our recent conference.

Do you currently use AI within Mercer/Marsh and how do you see it applied in practice across investment strategies?

Among the strategies analysed by Mercer's 200 analysts worldwide, several asset‑management processes do indeed use AI engines. Some hedge funds and quantitative managers are among the most advanced in integrating AI outputs into their investment decisions.

For investors, how much transparency and understanding of a model is realistically required before allocating capital and do you think they would they be willing to trade some performance for greater explainability?

As a preamble, note that this depends on each investor. Today, many of us use AI in our daily lives while accepting the limits of our understanding. In my opinion, transparency is key. Both the limitations and the strengths must be made explicit. That does not mean that the model must be fully understood or that performance can be completely explained. AI can increase discomfort, but some hedge funds are difficult to penetrate, as are some "traditional" quantitative processes.

From your perspective at Mercer/Marsh, what are the biggest barriers preventing broader adoption of AI-driven investment strategies today?

An increasing number of asset managers – particularly quantitative firms and hedge funds – are already developing investment strategies that leverage AI models. Others are integrating AI tools into their daily workflows to boost efficiency. Adoption is strong across the asset management industry, reflecting its broader uptake in everyday life. Key barriers include mastering the technology and attracting (and retaining) the necessary talent. Ultimately, as with any manager, performance remains the final arbiter.

What developments or enhancements would make institutional investors more comfortable allocating capital to AI-driven strategies?

I'm not sure. Reluctance to allocate to AI‑driven strategies may stem from a conservatism bias or from an illusion of control & knowledge (regarding non-AI strategies).

Looking ahead, how do you see AI evolving in investment management over the next 4 to 5 years?

Just as in trading – where algorithms increasingly compete against other algorithms¹, asset management has progressively embraced quantitative approaches. AI is the next evolution of this trend and is likely to continue expanding. Research shows that AI can create value², yet combining it with human analysts remains essential in certain changing market environments³. The pursuit of market inefficiencies remains a powerful driver; according to Andrew Lo’s Adaptive Market Hypothesis, AI can assist portfolio managers in this quest by enabling them to process larger volumes of data more quickly, moving efficiently from one identified inefficiency to the next.