Science fiction a decade ago, part of our daily life now: digital assistants on our smartphones, robot vacuum cleaners, self-driving cars… artificial intelligence is everywhere and its fields of application seem infinite, particularly in the asset management industry.
Artificial intelligence is defined as the ability of a digital computer or computer-controlled robot to perform tasks commonly associated with intelligent beings. The term is frequently applied to the project of developing systems endowed with the intellectual processes characteristic of humans, such as the ability to reason, discover meaning, generalize, or learn from past experience (Encyclopaedia Britannica). The infographic illustrates each technology related to artificial intelligence, each with its own level of maturity (don’t pay too much attention to the mentioned readiness as this chart is based on data from 2016 and 2017).
Let’s take for example Goldman Sachs CORE equity funds. The range, which encompasses big data analysis and natural language processing in its investment process,has collected more than €4 billion last year. This, however, is not an isolated phenomenon as dozens of other artificial intelligence-powered funds have been launched over the last couple of years.
Some asset managers may be at the leading edge of artificial intelligence implementation, but their clients – asset owners – may be quite advanced as well.
Almost 70% of institutional investors are currently already using, implementing or exploring artificial intelligence, according to the 11th edition of Fidelity’s Global Institutional Investor Survey:
- 10% of surveyed investors have already fully integrated AI into their decision making process,
- 24% are currently testing it,
- 35% are exploring its capabilities,
- and only 31% have no immediate plans to consider.
An interesting finding of the survey is the fact that AI adoption rate is relatively homogeneous across regions, with two notable exceptions:
- Tech-savvy Japan,with 38% of institutional investors having already integrated AI,
- Much more reluctant USA, with 75% having no plan regarding AI.
While these are still the early days of AI implementation, a majority of institutions which have started using it is already seeing the benefits:
As early as 2025, 60% of surveyed investors expect that AI will likely augment traditional investment jobs, while 53% expect it will replace many of them: 63% of European (ex-UK), 70% of Asian (ex-Japan) and 80% of Japanese investors think that investment allocations will be made without human contact.
The data suggests asset managers will need to carefully re-evaluate their position in the industry as technology has the potential to edge them out of their traditional roles.
AI is expected to be used to achieve several purposes, but optimizing asset allocation model and monitoring performance and risk come first:
To illustrate this in a concrete manner, let’s review two examples of artificial intelligence at work at some of the largest pension funds.
GPIF: AI to detect drifts in external managers’ investment style
Japan’s Government Pension Investment Fund, which manages $1.35tn of assets – making it the largest pension fund in the world –outsources a large portion to external asset managers. Over the past years, the fund has been struggling with under performance on most asset classes (negative excess returns), albeit bearing unsustainable increase in fees.
To address these issues, both a new fee structure (a lower base rate plus a performance fee rewarding excess returns) and a data-driven monitoring of external asset managers were introduced last year.
GPIF has partnered with Sony Computer Science Laboratories to build a proof-of-concept AI system. The system is based on a deep learning neural network designed and trained to detect eight investment styles from trading behavior data: in addition to external scenarios such as the market environment and corporate performance trends, descriptions of the stocks that constitute the fund (market capitalization at each point of time and unrealized loss or gain) and daily trading actions are input as time-series, and an output is a vector representing the style of the fund manager.
The system has been able to successfully detect the investment styles and drifts of each tested fund manager, as well as the convergence between fund managers with supposedly different styles.
The GPIF-Sony CSL research also points out that this method of evaluating manager is superior to some other current analytical tools, such as Barra model, which evaluate investment styles by examining changes in return, while the new system analyses the funds’ behavior itself, which makes it possible to detect style drift earlier and more directly.
Consistency checks between declarative descriptions and actual behavior, which have been done qualitatively, relying only on interviews so far, can be now conducted in an evidence-based manner. By removing some arbitrariness and subjectivity, the AI system would enable GPIF to better select and monitor fund managers based on stringent analysis.
According to GPIF, that can lead to significant improvement of the quality of interaction between GPIF and fund managers and of its investment practice, while managing risk properly.
APG: AI to identify best-suited companies to address Sustainable Development Goals
APG, the Dutch pension fund with €470bn of AuM, acquired Deloitte’s data analysis team of 13 people in 2018. The team continued as a separate business unit known as Entis. In one year, Entis categorized 10,000 listed companies around the world in a Sustainable Development Investment classification system, based on the United Nation’s seventeen Sustainable Development Goals.
How sustainable are their products, and can they be considered a Sustainable Development Investment on this basis? To find out, Entis scanned 400,000 documents (annual reports, websites, and Chamber of Commerce registration details) related to these 10,000 companies using artificial intelligence and machine natural language processing.
Regarding, for instance, the “affordable and clean energy” goal, algorithms are able to analyze all sources for key words like “solar panels”, “carbon emissions”, and “climate change”. This information is then compared to sales and revenue figures. This allows to determine the extent to which the revenue from a company’s products and services contributes to the SDGs (and consequently to detect those having green washing practices).
Looking for the “Teslas of tomorrow”, APG’s Director of Quantitative Investment explained they ended up finding 1,320 of them thanks to Entis system, much more than they had initially imagined.
Currently, APG is investigating how the patterns identified by Entis could improve the return of their investment strategy and whether they want to use this information to manage their €50bn-quantitative equities portfolio.
Insurance companies: AI to disrupt value chains and business models
Artificial intelligence is at the heart of the development strategies for more and more asset owners, as expected benefits go far beyond the sole field of portfolio and risk management. Additional examples are provided below, for an insurance company:
- Behavioral policy pricing: safer drivers to pay less for auto insurance and people with healthier lifestyles to pay less for health insurance, thanks to data provided by car’s or smart watch’s sensors.
- Customer experience and coverage personalization: in 2017, insurtech Neosurance and AXA Italy launched a “instant insurance”, that can be purchased in 10 seconds. The product, a travel insurance, is integrated into a concierge app and pushed through customer depending on his or her geographic data.
- Automated claims investigation and settlement: Nordea has started implementing computer vision, text analytics, machine learning and robotic process automation to deal with disability claims from life insurance customers, in order to speed up claim investigations (75 days on average). Nordea expects that two third of the cases will be handled automatically.
Conclusion
As we have seen, artificial intelligence broadens the scope of possibilities for asset owners in multiple fields, such as portfolio management, fund provider selection, distribution, sales, and client satisfaction (and we haven’t even mentioned blockchain technology – combining AI with blockchain should decuple these possibilities). It also raises the bar higher for asset managers.
Artificial intelligence comes with its benefits, but also with its drawbacks (cost of data scientists and infrastructure, redundancy, personal data protection…), nevertheless, how not to imagine it will continue to further disrupt our industry, from both asset owners’ and asset managers’ perspectives?
For asset managers, the artificial intelligence revolution demands a huge cultural shift, leading to a re-allocation and re-prioritisation of resources, and ultimately to an organisational change. But as it is becoming a key business driver, taking this road is, more than ever, a matter of survival.
Author: Eric Ancrenaz - April 2019
