Following our latest DigiBook article on digitalization impacts for funds selection, we carry on a cycle of articles diving into each dimension of the AM value chain. In this edition we cover the analysis of AI impacts on the research field.
AI deeply affects the asset management industry
Big data, machine learning, blockchain, Artificial Intelligence ... new technologies have a deep impact on each and every link of the asset management value chain. US consultancy Opimas estimates that 230,000 jobs in capital markets and more than 90,000 in the global fund management industry will disappear over the next seven years as a result of firms’ rapidly growing use of AI.
And the investors haven’t missed it: according to Venture Scanner 1,907 AI startups have raised $21.2 billion in venture capital funds before even counting investments made by financial companies and technology giants.
Why such a keen interest? Because AI can improve productivity, cut costs and provide competitive advantage.
According to leading research firm Tractica AI revenues will reach $59.8 billion worldwide by 2025 (up from $1.4 billion in 2016). Statista makes a similar forecast with $38 billion revenue. A major issue!
An increasing pressure on research costs
Regulatory requirements (MiFID II) are putting equity investment research under the spotlight. Asset managers are pushed to both improve the quality and reduce the costs of this activity (5 to 20 basis points of AuM according to the CFA Institute). Indeed, delivering alpha to clients has become crucial at a time when they are increasingly shifting to passive strategies.
The abilities of AI could meet these goals because investment research is its natural field. AI is particularly efficient when there are too many or too complex data (diversified, unstructured, etc.) for a human to analyse.
Let's dive into the concrete applications of AI in macro-economic, fundamental and quantitative research.
Macro-economic & financial market analysis: AI provides extended capabilities
Forecasting economic and financial market trends is the essence of asset management. With AI, and especially deep learning, algorithms can use big data to analyze in an automatic and self-learning manner the key factors impacting the market. Technology thus allows to set up new added-value processes which can’t be operated by humans. Examples are plentiful: websites and social networks screening, exploitation of satellite data, financial modelling of multiple markets patterns, analyse of e-commerce sales data, etc.
The types of analysis could thus be enlarged with AI:
- Market analysis modelling stock markets’ behavior and trends
- Behavior analysis exploring consumer or economic development data
- Feelings analysis scanning articles and social networks to give market sentiment to a company or a market
Many providers have emerged, producing indicators for traders or portfolio managers. Some competitors have already integrated these analyses in their processes such as:
AXA IM uses data from MKT MediaStats (analysis of large sets of consumer data which reflect their behavior in the real economy) and PriceStats (scan of online sales data on 5 million items sold by hundreds of retailers in 70+ countries) to measure daily inflation, in order to fine-tune and accelerate its investment decision-making process.
La Française AM works with the Urban Morphology Complex Systems Institute, which relies on big data analysis to identify priority real estate investment areas.
Fundamental analysis: new data are accessible
New technologies like Natural Language Processing (NLP) can dramatically enhance efficiency of traditional fundamental analysis by reviewing millions of documents and press by scanning every materials available on Internet (accounts, regulatory filings, conference minutes, articles, etc.).
To widen this classic investigation field, machines are now able to synthesize valuable information contained in unstructured data (tweets, internet searches, social media, satellite feed, etc.). It is even possible– through fragmented information – to detect changes in the reputation of a company (sentiment analysis) or early signals of the success/fail of a product.
Below an illustration from Goldman Sachs AM on how big data could extract investment opportunities.
BlackRock’s Scientific Active Equity (SAE) unit, like Principal Global Investors, deploys big data analytics to identify potential investment flags (conference calls releases, google satellite feed, internet searches ...). Last March, BlackRock announced a $30Bn shift from traditional stock picking to the SAE quantitative processes.
The British AMh created a Data Insight Unit, with 16 data scientists and AI experts, to change the way its PMs make decisions with new data (tweets…). Schroders states that its processing cluster is 527 times more powerful than Deep Blue, the chess-playing IBM supercomputer!
IBM's Watson supercomputer was hired to help run an actively-managed ETF(“Equbot with Watson AI Total US ETF”) and pick stocks. Equbot uses Watson AI to perform a fundamental analysis on U.S.-listed stocks and real estate investment trusts based on up to 10 years of historical data and then apply that analysis to recent economic and news data.
Quantitative analysis: from hedge funds to mutual funds
Hedge funds were the first to use AI in investment management. It is estimated that 40% of hedge funds launched in 2015 used AI algorithms for making investment decisions. Two Sigma, Bridgewater Associates, Renaissance Technology or Point72 are the flagships of the rise of AI-managed hedge funds.
More recently, traditional financial institutions have also implemented AI in their investment process, identifying patterns or introducing active management in passive strategy.
Goldman Sachs appears as a pioneer. In 2014, they invested and began installing an AI-driven trading platform called Kensho. Now, the quantitative-investment strategies division at Goldman Sachs uses language processing driven by machine-learning going through thousands of analysts’ reports to support stock picking decisions.
Goldman Sachs AM as shown in the infographic above also uses AI and big data for its quantitative equity range such as its GS CORE Equity Portfolios which is the best selling actively managed Global Equity fund this year (€2.3bn net sales for a total AuM of €3.2bn)
JPMorgan has deployed a machine learning to drive predictive recommendations. “Emerging Opportunities Engine” identifies clients who are best positioned to buy or sell stocks, through an automated analysis of their current financial positions, of current market conditions and issuance history. JPMorgan intends to expand the engine to debt markets.
AI and machine learning applications show substantial promise and could lead to sales success as shown by Goldman Sachs AM. However, they require change management because of their huge impacts on human resource management (specific skills needed). The consequences on the job market could be significant as Goldman Sachs showed by reducing drastically its number of traders (today, U.S. cash equities trading has only two traders, machines are doing the rest). Otherwise, AI requires new talents such as data scientists or machine-learning experts (see DigiBook #8 article on Data Ninjas). Two Sigma and its peers demonstrate this as they are actively recruiting such specialists.
Author: Pascal Buisson - November 2017
