The advent of artificial intelligence (AI) into fund performance analysis will help asset owners quantitatively determine which funds generate excess returns through astute performance, pay them commensurately and more effectively analyse their ongoing investment approaches, believe fund managers and technology experts.

To date, the most public use of AI to analyse fund management performance has been conducted by Japan’s Government Pension Investment Fund of Japan (GPIF). As reported, it appointed Sony Computer Science Laboratories (Sony CSL) to decide which parts of the investment process can most effectively use AI.

Sony CSL’s efforts have to date mostly focused around analysing data points to assess how much a set of fund managers mandated by GPIF tended to ‘drift’ from the investment mandates they had been hired to pursue.

This approach offers a potentially quantitative means to analyse the investment approach of fund houses. Up to now, the evaluation of the risk/return ratio of a fund has basically been a matter of back-testing historical data. However, once an investor can obtain a predictive model of a fund manager's behaviour, it is possible to simulate the predictive evaluation of a fund manager's trading actions under arbitrary scenarios, as well as their predicted performance and risk characteristics.

The next iteration of the AI programme might also allow asset owners to detect which funds have outperformed through skill and which were lucky, benefiting from circumstances beyond their control.  

Paul Sandhu, head of multi-asset quant solutions at BNP Paribas Asset Management in Hong Kong said that could change how asset owners pick funds. 

Paul Sandhu,
BNP Paribas Asset Management

“If you take a market downturn as an example, a human is going to perform based on what they think could happen," he told AsianInvestor. "They might say this is a momentary downturn and you might want to buy more equity, or sit tight for a while.

"It’s difficult to hold them to a certain decision because it can depend on certain factors. The difference with an AI or a model approach is there is no ambiguity about. If this happens, it will do this, and that is going to give us this result. A lot of people get comfort in that.”


Currently, conventional fund platforms such as Barra or Aladdin evaluate investment styles by examining changes in return by factor and the sensitivity to each factor. And the evaluation of the risk/return ratio of a fund has basically been a matter of back-testing historical data.

But Sony CSL said the system it developed using GPIF data directly analyses the funds' behaviour itself, making it possible to detect style drift earlier and more directly. And once an asset owner can obtain a predictive model of a fund manager's behaviour, it becomes possible to use simulation for predictive evaluation of a fund manager's trading actions under arbitrary scenarios, as well as the prediction of performance and risk characteristics.

“[The programme] should make it possible to carry out genuine forward-looking risk/return evaluation and stress testing, rather than relying only on past records, which we believe would lead to the construction of a more robust manager structure,” the company said.

Takao Tajiri, project leader for the GPIF’s AI project at Sony CSL, told AsianInvestor that Sony CSL is now widening the prototypes data set, attempting to quantify how much of fund performance is down to skill versus market movements.

Takao Tajiri, Sony CSL

It will then combine “these components into an integrated system in order to provide an additional new way of evaluating fund managers, which will augment GPIF officers' business processes”.


While Sony CSL’s technology has been designed for the vast assets of GPIF, Tajiri believes it could prove useful for smaller asset owners too, noting that pension funds of all sizes face similar challenges when it comes to identifying good external managers.

To meet this, he and his team have proposed that global asset owners establish a data consortium that pools all data about their investment activities. That would accelerate and improve the prototype AI’s learning ability, and make it applicable for other research too.

Practically speaking, AI could also help smaller pension funds overcome a dearth of qualified investment professionals. “Pension funds with relatively smaller assets or those located in provincial cities in Japan are strongly affected by [the country’s] depopulation, so the recruitment market has been tighter than ever,” said Tajiri. “We expect our new AI solution could enhance their monitoring and assessment capabilities.”

Institutional investors in Japan and across the world will be likely to follow the impact of GPIF and Sony CSL’s AI prototype with great interest. The results could help them to better identify fund houses that better represent their investment preferences and identify alpha and beta due to judgement rather than luck. That could mean the writing is on the wall for underperforming active managers.

As Tajiri said, “When asset management companies recognise that GPIF has the ability to independently analyse their investment styles and intends to continue development of even more advanced technology, they will recognise that they cannot justify their results with only qualitative explanations.

“Some funds will fail to develop and maintain systems that efficiently generate excess returns, and will be forced to either leave the market, or move towards index trading.”

The latest edition of AsianInvestor magazine contains a special report on technology in investment management.