Artificial intelligence in fund management
Future of industry outlined in a survey across nearly 300 fixed income professionals
9 Jan 2019 | Asset Benchmark Research
FOR the second year in a row, Asset Benchmark Research (ABR) contacted 300 buyside and sellside fixed income professionals across the region to assess attitudes towards artificial intelligence (AI) and its application in the market. Far from fear over how AI technologies might make some of them redundant, the study reveals a long wish list among investors and sellside individuals that has been met with insufficient real-world application.
Nearly half of surveyed investors believe AI will play a substantial role in fund management, analytics and trading in the future. One-third comment that it will assume a limited role only. One in ten investors shrug at the idea altogether, forecasting that AI will never assume any role in fixed income fund management.
Six percent of respondents say that they have already felt the impact of AI in the industry. Some predict that AI will play a larger role in obtaining quick analysis of information, trends and patterns without any unconscious bias. Nearly a fifth of respondents foresee the possibility of AI application in data analysis and visualization.
“AI can gather all fundamental information – rate and FX movements, as well as valuations to see patterns and impacts of changes in these factors. Many investors are trying to avoid the bias or irrationality. We can see fund managers applying robotic investment decision making in other asset types,” says an investor of a Thailand-based investment house.
Fund managers also understand AI’s limitations. “AI will definitely have a role in the future, but will be limited to data analysis, recommendations, and forecasts. The decision making will always be done by the traders. AI can’t have a feel for the market unlike traders do,” comments a fixed income investor based in the Philippines.
Investors are also confident that they are better than AI in grasping the direction of markets and factoring in sentiment. “AI has certain limitations, for example, market sentiment is difficult to quantify in data form. Past track record differs from the current scenario,” says Faiz Gany, a portfolio manager at Etiqa Takaful Berhad in Malaysia.
A large group of investors also argue that the lack of depth, liquidity, and volatility in local currency markets will prevent AI technologies from being fully applied in the fund management industry.
“AI would only work in areas where it takes humans a few seconds to come up with decisions, i.e. capturing market anomalies and short-term trade ideas. But for those which take us days or weeks to decide, i.e. strategic decision, AI won’t do the job accurately. I would rather put higher emphasis on data science, which is key to a fund manager’s decision making,” says Arsa Indaravijaya, head of Investment Strategy of Government Pension Fund in Thailand.
Luc Froehlich, APAC head of investment director at Fidelity International, points out the complex nature of accountability in the event when an AI tool leads to investment losses. Speaking at the recent 13th Asian Bond Market Summit held in Singapore, he noted: “Currently the buzz words are ‘deep learning’ or ‘artificial neural network’. While it’s an exciting field, which could in the future help generate attractive investment performances, I see some challenges there. With this type of approach it can be extremely difficult to establish a linkage between the input and the output [that is the information fed to a machine and the performance of a fund]. Concretely that means that in certain instances, it might be barely possible to explain to an investor how the performance of his fund was generated. Basically, ‘computer says no’.”
Thinking two steps ahead
Similar to their peers in the buyside, most sellside individuals are convinced that AI can bring benefits to their day-to-day tasks. Based on a poll, an overwhelming majority of respondents (95%) believe AI will play at least some role in the fund management industry.
Francis Pong, credit trader at DZ Bank, has a specific scenario in mind in which AI could help manage portfolios in volatile market conditions. “It would be nice if spreads on ALLQ be updated automatically based on last traded price. We currently have 600-700 IG bonds to update every day. When the market is volatile, we are unable to cope. Clients complain that prices are unrealistic, which is unavoidable as most trading teams are lean,” he says.
Indeed, somewhat surprisingly, most sellside individuals name the menial tasks of their jobs first when asked about AI application in their roles. Manual booking of trading tickets tops the list of tasks that respondents hope AI can handle in the future, followed by generic data analysis, and other administrative work such as trade settlement and execution.
One interpretation of these results is that people enjoy challenging tasks, strive for excellence, but dislike being tied to tedious manual routine. “Actually, automation has limited benefits to true credit research,” says Owen Gallimore, economist at ANZ. “The minutiae are where a good credit analyst can find warning signals, be in the notes or financial statements of an earnings announcement, or in conversations on issuer and investor trips.”
The fact that few respondents look to AI solutions for KYC (know your customer) and client on-boarding procedures as well as answering client queries suggests that machine intelligence cannot replace the utmost importance of trust between a salesperson and his client.
William Chao, head of Taiwan credit sales at Standard Chartered adds: “Rather than replacing sales functions, AI can assist a salesperson to do a better job. First, AI can handle all the trivial work, such as pre-deal checks, trade details, recording accurately and efficiently before every transaction. By doing so, AI can help largely reduce administrative mistakes and operational risks. Sales will have more client-facing time and focus on driving more revenues instead of being fettered by increasingly demanding regulations.”
“Second, AI has the ability to memorize the preferences of different clients, which can help the salesperson to dig out the potential demands of the clients and create new business ideas that the salesperson might not come up with before,” he adds.
They say a company is no longer worth buying if the taxi driver calls its shares the next growth story. Bitcoin investors can certainly relate to this. Considering the market buzz around AI, one might be led to believe that AI is yet another hype.
Yet those introducing AI in the fixed income market have clear views on the challenges mere coding can address and the roles that will remain firmly under their purview. They are also clear of the fact that AI is here to stay.