The arrival of artificial intelligence (AI) is a silent revolution. It will change all sectors, including asset management. Investors are challenged to grasp the nature of this new reality and how artificial intelligence can participate in creation of added-value for listed companies.
AI and machine learning1 are the top priorities for most asset managers worldwide2. In the financial services sector, AI is caricatured as a substitute for human employment. Stating this would be particularly short-sighted as in such a complex activity as asset management, AI is still in its infancy.
In the past, financial markets mainly operated based on structured data such as growth rates or financial indicators. However, 80 percent of the data currently available is unstructured data, such as press articles or tweets. For asset managers, this data is indispensable as it allows them to capture the market sentiment towards each company. Besides language barriers, the sheer volume of data is a problem. It sets strict quantitative limits to what can be analyzed at all - so far.
Natural Language Processing (NLP)3, a special AI discipline, makes it possible to overcome these limits. Let's take the example of a global equity management: of the approximately 4,000 companies that compose the investment universe, the traditional fund manager may know a thousand. He will therefore not be able to guarantee his investor that he has explored all the possibilities in his universe. But thanks to NLP and machine learning, he can now access all data concerning these 4,000 companies: With this tool, he can scan his entire universe and map it in the most intelligent way. This effectively eliminates distortion of attention: We put a stop to the trend of a decision-making based on a limited number of factors familiar to them or of a particular concern.
This expansion of a fund manager’s universe of potential knowledge and the acceleration of access to relevant information make it possible to rapidly multiply specific investment strategies such as "Chinese consumption" or "ageing of the population" in the sense of an expanded selection without substantial investments. In other words, the tools derived from the AI have great potential for tailor-made asset management.
If the machine is able to find and select relevant information for asset management, what role do humans still play? A very central one, as the fund manager remains involved and calls the shots at the beginning, during and at the end of the process leading to the investment decision.
In the beginning it is the fund manager who prepares an investment thesis: For example, companies engaged in the AI sector should outperform in the medium to long term. Based on this premise, the fund manager builds up a universe and defines a series of criteria that will allow him to select what are probably the most efficient companies.
After completing this work, they "translate" business intelligence experts, the famous data scientists, into machine language. The goal is to develop so-called intelligent algorithms that are able to generate relevant data for the investment on their own.
The algorithm must be permanently trained and, if necessary, corrected if it delivers results irrelevant for the management. This work is the result of an ongoing dialogue between fund managers and IT specialists. Finally, before the investment, it is still the responsibility of the fund manager to ensure that no extraordinary factors have changed his investment theory. In the event of an aircraft crash, for instance, it is up to the fund manager to determine the significance of the impact on the aircraft manufacturer’s economic solidity.
What impact will AI Asset Management have on institutional investors? It implies a kind of philosophical choice between active management, quantitative management, index management and hybrid strategies based on AI.
The choice depends essentially on the investor’s needs and beliefs. He must be aware that machine learning is probably the most powerful tool to free themselves from indices. While none of the outlined possibilities is a miracle cure, it may be promising to combine them to exploit the specific characteristics of others: hybrid strategies for example, for their predictive qualities and quantitative management for its systematic approach, which allows tendencies such as momentum to be fully exploited.
For investors with strict governance requirements, AI-based asset management can be an advantage in terms of credibility. Thanks to AI, all research, including unstructured data, is carried out in order to guarantee investors the best possible results.
In terms of performance, it is too early to draw conclusions. However, accelerating access to information should increase the chances of achieving above-average results. In addition, as AI is not yet widely used in asset management, it offers a competitive advantage. This could remain so because the algorithms are not interchangeable. There are as many ways to use them as there are human brains.
1 Machine Learning is a type of AI that allows computers to learn without being explicitly programmed beforehand.
2 Statement of the consulting firm Deloitte.
3 Natural Language Processing (NLP): A field of computer science, artificial intelligence and computational linguistics that deals with the interactions between computer and human (natural) languages, including the programming of computers to process large quantities of natural language bodies.
Source : translation into English of the article from Prévoyance Professionnelle Suisse | Schweizer Personalvorsorge, November, p.100-101