What if AI replaces human investors? | Prof Byoung-hyoun Hwang
In Episode 2 of Nanyang Business School’s What If series, which explores thought-provoking questions in the age of AI, Professor Byoung-Hyoun Hwang – President’s Chair Professor in Finance at NTU and Head, Division of Finance at NBS – explains why fiduciary accountability, markets that react to their own predictions, and the need for creative thinking will keep human investors indispensable in an algorithmic world.
Interested in partnering NBS on your business transformation? Reach out to us at: nbscomms@ntu.edu.sg
Demystifying AI's Role in Modern Asset Management
AI will not make human investors obsolete. The reason is structural. Ultimately, it will always be humans providing the capital. It is our money, and it is fundamentally up to us to decide our financial goals, our acceptable risks, and the constraints we face. That is something AI will never be able to take away.
There is a second, equally fundamental reason. The current legal and regulatory regime simply does not allow for humans, as an accountable fiduciary party, to be entirely removed. There will need to be humans supervising what AI is doing and intervening when required. While the role of human investors will change, they will always remain an indispensable part of the system.
This realistic perspective is grounded in the deep academic and market expertise of Professor Byoung-Hyoun Hwang, Professor of Finance at Nanyang Business School (NBS), whose research focuses on behavioural and social finance, and increasingly, the intersection of AI and financial markets.
A Century of Finance: From Pencil to AI Agent
To understand how this role is evolving, it is worth tracing the historical trajectory of investment analysis:
1. The Traditional Era (95% Human, 5% Machine)
Historically, when deciding whether to allocate capital to a company, an investor had to estimate its fair value by hand. This meant interviewing potential customers, consulting industry experts, and projecting the company's future cash flows – then working out the price at which it should trade relative to its actual current price. It’s a methodology still taught in introductory finance classes today because it illustrates key financial concepts. Doing this required little more than paper, a pencil, and a calculator. Prof Hwang frames this era as 95% human and 5% machine, defined by the physical calculator.
2. The Quantitative Era (Rules-Based Trading)
The advent of computers unlocked vast repositories of data, enabling the systematic search for statistical patterns. This birthed rules-based trading.
Consider a simple example: the NYSE and NASDAQ list roughly 5,600 stocks. An investor might hypothesise that stocks experiencing extreme price surges on a given day will revert the next day, while those experiencing sharp declines will bounce back.
While some price movements are driven by fundamental news, others are caused by sheer buying pressure from large institutional funds executing high-volume orders. To deploy capital, these funds must acquire millions of shares, temporarily inflating prices due to limited supply. Conversely, liquidating holdings exerts downward pressure, depressing prices without any change in company fundamentals.
With computers, investors could backtest these hypotheses against decades of historical data to verify profitability before automating the execution.
3. The Scale of Algorithmic Assets
Over the past few decades, quantitative trading has expanded dramatically:
- High-Frequency Trading (HFT): Operating at the millisecond level, HFT is now estimated to account for about half of all equity trading volume in the United States.
- Smart Beta and Long-Short Quant Funds: For longer investment horizons, automated systems can sort stocks every quarter by metrics such as price-to-earnings (P/E) ratios, automatically buying undervalued assets and shorting overvalued ones. Today, a tiny team of managers with a computer can easily oversee a multi-billion fund. This quantitative approach commands over $3 trillion in assets under management (AUM) worldwide.
What is Different About AI This Time?
While technology-driven investing is not new, modern AI represents a paradigm shift rather than an incremental step.
In high-frequency and smart-beta environments, it was humans who had to come up with the specific ideas and hypotheses themselves and later program the machine to test them. With AI, we can design autonomous agents and instruct them to develop and test ideas independently.
Under this framework:
- Agent 1 (The Creator) could synthesise the data and propose and test trading strategies.
- Agent 2 (The Validator) could audit the first agent's program to ensure the code is sound and its underlying assumptions are reasonable.
- Agent 3 (The Devil's Advocate) could challenge the strategy, actively searching for structural weaknesses.
These agents could collaborate and iterate autonomously. While the human investor retains ultimate authority over capital allocation, the machine has been promoted from a basic research assistant to something closer to an autonomous postdoc or research associate.
Three Limits Worth Taking Seriously
These breakthroughs are remarkable, but three limitations deserve serious attention:
1. When Everyone Uses the Same Models
Many institutional investors could be relying on the same foundational models – such as Gemini, Claude, and GPT – using similar prompt structures. When thousands of fund managers run related models on the same datasets, they inevitably generate highly correlated strategy recommendations. In financial markets, a trading strategy loses its profitability the moment it is widely adopted. Alpha relies on unique insight and information.
2. When Predictions Change the Outcome
If an AI model predicts rain, carrying an umbrella does not alter the weather. However, if an AI model predicts a severe influenza outbreak, prompting hospitals to prepare, and citizens react by masking up and staying cautious, the outbreak may never materialise.
This is the principle of endogeneity: human behaviour reacts to predictions, which in turn alters the outcome. If an AI model identifies an attractive trading strategy and the market rushes to exploit it, the opportunity is instantly arbitraged away. Even worse, it could lead to bubbles and crashes. Over-relying on predictive models that ignore their own market impact is exceptionally risky.
3. The Accountability Gap
Finance is built on fiduciary duty. If a pension fund manager – such as one representing a teachers' union – loses money shorting technology stocks, they can explain their thesis to the trustees: “I believed the market was overvalued, I took a calculated risk, and I was wrong.”
If that same manager delegates all capital allocation to an autonomous AI agent and suffers catastrophic losses, the honest explanation is: “I have no idea why the machine made those trades. I don't even know what this thing is trading on.” For institutional asset managers, this lack of auditability is legally, politically, and professionally untenable.
Beware the Tipping Point
Because market anomalies are quickly discovered and eroded by competitors, the financial sector rewards speed. Being the first to identify and trade on an anomaly is “gold.”
Singapore is actively positioning itself as a first mover in financial AI applications. At this stage of adoption, the strategic priority is expanding market share rather than maximising profitability.
However, investors must remain highly agile. Quantitative strategies will generate diminishing abnormal returns as they are discovered by other investors. AI offers new opportunities by discovering novel patterns and generating trading ideas that humans may overlook, but at scale also introduces systemic risk. If many funds rely on similar AI-generated strategies, markets could experience increasingly correlated trading, amplifying volatility and reducing stability.
This dynamic creates what Prof Hwang describes as a “tipping point” in AI adoption. Investors should know that there is such a tipping point and thus act fast while growing cautious as it approaches.
The Luddite Fallacy, Revisited
The anxiety regarding career obsolescence among university students is understandable but historically not unprecedented.
During the 1970s and 1980s, the introduction of automated teller machines (ATMs) led to widespread predictions that bank tellers would be eliminated. Instead, because ATMs lowered the operating cost of running a bank branch, banks opened significantly more branches. Consequently, the total number of bank tellers actually rose.
However, their job description shifted from manual cash handling and deposits to loan appraisals and advisory services.
A similar shift could be happening in finance. The junior analyst's role is transitioning away from manual data work towards high-level strategic evaluation.
Inside the NBS Classroom
At Nanyang Business School, this transition has directly influenced our curriculum. In the undergraduate course Equity Investing with Big Data, for instance, the syllabus has shifted from manual programming to agentic oversight:
- The Old Way: Students spent weeks writing manual code to clean datasets and backtest strategies.
- The New Way: Using advanced code-generation environments like Claude Code, students set up AI agents. The agent parses the data, suggests trading strategies, and writes the code, while students focus on critiquing the economic intuition and structural validity of its ideas.
This hands-on training is supported by direct collaborations with hedge funds. Students with strong economic intuition can now execute complex quantitative analyses without being held back by coding syntax. The educational focus is shifting from technical execution to creative, first-principles thinking.
Our continuous conversations with elite peer institutions in the United States show that NBS is ahead of the curve in integrating operational AI budgets and API tokens directly into the student learning experience.
The Human Edge: Out-of-the-Box Thinking
The ultimate competitive advantage in an AI-saturated market does not lie in raw computing power; it lies in the humanities, philosophy, and creative problem-solving.
Just as AI-generated videos and templated admission essays quickly become repetitive and derivative, purely AI-driven investment strategies lack the creative breakthroughs needed to find truly unique opportunities.
To prepare the next generation, our educational models must prioritise creative play and multidisciplinary thinking over rote-learning. To remain resilient in a highly automated financial landscape, we must teach future leaders how to ask the right questions with rigorous “good taste.”
