AI Transforms Retail Investors into 'DIY Hedge Funds'

Artificial intelligence (AI) is breaking down the boundaries between Wall Street and retail investors. This comes with the emergence of AI agents that continuously monitor global stock and virtual asset markets around the clock, read news and public disclosures, filter investment stocks, and even execute orders when pre-set conditions are met.

There is also an increasing number of cases where individuals directly build automated investment systems that could previously only be established at hedge funds and large financial companies. This is thanks to the recent spread of 'vibecoding.' Vibecoding is a development method where AI writes the program on your behalf when you describe the desired functions and operation mechanisms in natural language.

According to major foreign media and the financial investment industry on the 3rd, individual investing utilizing AI has gone beyond the stage of simply asking a chatbot for promising stocks. From programs that automatically collect disclosures, news, and stock price data to stock screeners, investment research assistants, backtesting systems, and automated trading agents, individuals are directly building the investment tools they need.

AI Traders Scanning the Market 24 Hours a Day... The Era of Retail Investors as 'One-Person Hedge Funds'

In June, the U.S. economic media outlet Business Insider focused on the phenomenon of retail investors using AI to create their own stock screeners, research assistants, and trading agents. Investors build their own systems by explaining their investment strategies, trading signals, and entry and exit conditions to the AI, and connecting real-time market data. The article also introduced cases of retail investors who started with nothing but a computer and AI services like Claude and ChatGPT.

The base of investors utilizing AI is also rapidly expanding. According to a global retail investor survey by eToro reported by Reuters, 13% of the 11,000 respondents said they are already using ChatGPT or Gemini for stock selection or portfolio changes. About half stated they are willing to use AI investment tools in the future.

Alex Svanevik, CEO of the blockchain data analytics firm Nansen, predicted that within the next two years, the number of AI trading agents active in the market could exceed that of human traders. Nansen is also expanding its domain from its traditional data analysis-focused business to trade execution through AI agents.

However, it is difficult to conclude that AI can yield superior investment returns compared to humans. Nansen is also conducting backtests and mock investments before entrusting real funds to autonomous agents. CEO Svanevik revealed a case from their own experiments where an AI agent generated a profit of $23 while incurring $700 in AI model usage fees. This means there is still a significant gap between technical feasibility and economic viability.

'One-Person Research Centers' Growing Faster Than Automated Trading

The core of the spread of AI investment agents is not limited to automated ordering. The fact that individuals can now create their own investment research infrastructure is considered a bigger change.

For example, an AI agent can collect corporate disclosures from the U.S. Securities and Exchange Commission (SEC) and South Korea's Data Analysis, Retrieval and Transfer System (DART), find changes in sales, operating profit, and guidance, and organize them into a table. By combining this with news, stock prices, and trading volume data, it is also possible to automatically filter out stocks with surging trading volumes or those that meet certain financial conditions.

In April, Maeil Business Newspaper reported on a case of directly building an agent using Claude Code to collect and organize data from DART and the Korea Exchange (KRX) statistical system. It was structured so that one agent collects data and another samples and verifies the results. A function to issue task instructions via Telegram on a mobile phone was also applied.

By connecting investment strategies and securities firm APIs to such a structure, a personal investment system that leads from data collection to stock discovery, investment judgment, backtesting, and order execution is completed. In the past, this required significant development costs and specialized personnel, but the barrier to entry has been greatly lowered as AI coding tools support everything from program creation to bug fixing.

Cases of retail investors in South Korea building their own quant bots through vibecoding are also emerging. In May, Seoul Economic Daily reported that office workers with no coding experience were using AI and securities firm APIs to create overseas stock investment programs and were sharing algorithms and bug fixing methods with other investors on GitHub.

Securities Firms Also Opening AI-Readable Investment Infrastructure

Domestic securities firms are also responding to the personal investment agent market.

Korea Investment & Securities has released open-source codes for investment strategy generation and backtesting utilizing its Open API. Users can combine various indicators, such as moving averages and momentum, to create buy and sell conditions and verify performance using historical data.

It has also launched a dedicated plugin that connects with major AI coding agents like Claude Code, Cursor, Codex, and Gemini CLI. If a user inputs, "Create a strategy that buys when the RSI is 30 or below," or "Backtest Samsung Electronics over the past year," the AI is structured to perform the strategy design and verification. Currently, it supports 10 basic strategies and 80 technical indicator combinations, and can also be linked to mock and real trading orders.

Toss Securities' newly released Open API attracted 55,000 people within just two weeks of receiving pre-applications. The Toss Securities API connects domestic and international stock quotes, account, and order information to external programs and can also be integrated with AI services like Claude Code and Codex. It also provides 'llms.txt,' an AI-dedicated document, so that users can instruct quote inquiries, account analysis, and orders in a conversational style without knowing any programming languages.

An "AI-Generated Strategy" Does Not Guarantee Profits

Just because building an investment system has become easier does not mean that a good strategy is automatically created. The investors interviewed by Business Insider also pointed out that AI does not invent winning strategies on its own, and that a basic understanding of the market and clear trading principles are necessary first.

Data accuracy is also an issue. General-purpose AI models may not have access to paid information or may incorrectly quote numbers and dates of corporate earnings. Errors in predicting the future by relying too heavily on past price trends can also occur. If incorrect news or manipulated data is inputted, there is a possibility that the AI's investment judgment itself could be distorted.

Security risks are more direct. Giving an AI agent order authorization over a brokerage account could lead to actual financial losses due to program errors or external attacks. This is why Korea Investment & Securities mandates user confirmation before a real order is placed and applies separate security features to prevent API keys and tokens from being exposed in the code.

The new era of investing opened by AI is far from 'an era where you entrust your money to AI and it automatically generates profits.' Rather, it is closer to an era where individuals directly possess the tools to collect data, discover stocks, and verify strategies just like institutional investors.

The technological gap between institutions, which used to monopolize information and systems, and individuals is rapidly narrowing. Now, the standard that determines investment performance is highly likely to shift from mere information access capabilities to what data is selected and what principles and safety devices are designed into the AI.