Axa using AI to boost manager performance

5 April 2017, by Robert Van Egghen. 

Axa Investment Managers is exploring the use of artificial intelligence to improve its investment decisions, including analysing the behaviour of its portfolio managers.

Global chief operating officer Joseph Pinto says the firm is partnering with financial technology start-up Essentia Analytics to examine its portfolio managers’ “behavioural patterns in investment terms”.

Mr Pinto says Axa IM will use past performance data to glean insights that can be used to improve individual managers’ future performance.

He believes being able to analyse managers’ behavioural patterns will also mean Axa IM can anticipate any future decisions a manager may make, which could help the firm gain a headstart against its competitors.

Essentia Analytics, which also works with Union Investment and BNY Mellon fixed income subsidiary Standish Mellon, analyses trade and holding data, as well as historical performance data.

It also supplements this data with managers’ own personal diaries, including details like how much sleep they had the night before or their current emotional state.

CEO Clare Flynn Levy told Ignites Europe that the feedback, which is given directly to the portfolio managers, acts as “an electronic wingman” to them.

Ms Flynn Levy says this gives portfolio managers the same level of insight into their performance as professional athletes have, enabling them to “capture every aspect of their performance and see how they can improve it”.

However, Mr Pinto cautions that “it is too early” to draw any solid conclusions from Axa IM’s use of this behavioural data.

Mr Pinto says developing solutions based on machine learning, a form of AI in which machines analyse vast amounts of data and learn for themselves, is one of Axa IM’s key business objectives.

“We are engaged in a deep dialogue with a number of fintech companies,” he says.

Mr Pinto says the firm is currently testing systems that utilise deep-learning techniques, a subset of machine learning that utilises algorithms modelled on the structure and function of the brain. However, he declined to comment on how these could be used.

Axa IM is already using machine learning to analyse economic and market data for investment insights, having partnered with State Street and MKT MediaStats last year.

MKT MediaStats analyses economic data from multiple media sources to determine market implications while State Street’s PriceStats inflation series provides a daily measure of inflation based on prices posted to public websites by hundreds of online retailers.

However, Mr Pinto says Axa IM does not intend “to replace portfolio managers” with robots as “we still need [the managers’] long-term view”.

He says the attraction of AI is that it “gives us an idea as to whether there is a better time to enter into the market”, meaning fintech firms “give us market signals while the long-term decision is still in the hands of portfolio managers”.

He says: “We are [operating] in a world that is more and more competitive, and the pressure [to reduce] fees is something we see every day.

“We need our managers to have [access to] more valuable information and analytics to make better decisions.”

Richard Peterson, CEO of MarketPsych, which provides data analytic services to asset managers in Europe and the US, says using data to improve managers’ performance will become increasingly important as the industry shifts away from active funds.

“Active portfolio management is declining. Because on average active managers underperform there’s a huge move towards quantitative investment,” he says.

Mr Peterson cites the example of BlackRock, which recently laid off around 40 employees in its active equity business.

However, he cautions that since “you can’t ask the computer why they made a decision, there’s a real comfort in having a manager who can say why they did what they did”.

Yet Mr Peterson also warns that too much analysis may in fact impede the decision-making process and weaken performance.

He says: “Writing a daily journal and other more intensive interventions interrupt the decision-making process.

“Some self awareness is important, but too much can force solutions that interrupt the flow of the decision process rather than benefiting it.”

Mr Pinto says Axa IM is also utilising big data to examine investor behaviour.

He says the firm has been examining the inflows and outflows from its mutual funds during major market events.

“What we’ve found is that the same customers behave in the same way every time,” he says.

Mr Pinto says that whenever a market trend occurs, such as a shift towards US equities, the same clients make the first move towards or away from that asset class, with a larger group of investors jumping on the bandwagon, before being followed by the same group who are always the slowest to react.

He says that while the project remains “in the beta stage” the data have already enabled Axa to “do some customer segmentation”, which can be “used for marketing purposes and enable us to better serve their needs”.

See how Essentia’s behavioral nudges and investment insights helped one active manager to improve his exit timing and position sizing, and generate significant incremental alpha for his investors.