Published on
For decades, quantitative trading was one of finance’s most exclusive fortresses. Behind the closed doors of specialised hedge funds such as Renaissance Technologies, Citadel and Two Sigma, teams of mathematicians, physicists and computer scientists built proprietary algorithms that could spot market inefficiencies in fractions of a second. With expensive server farms, servers placed right next to the exchange and exclusive data feeds, they played a different game from everyone else.
If you were an everyday investor, you brought a spreadsheet to a supercomputer fight.
Today, that fortress has been breached. A large share of trading volume on the major markets has been algorithmic for years, and the most important shift is no longer happening only on institutional trading floors. Thanks to large language models, machine learning and no-code platforms, techniques once reserved for professionals are within reach of anyone with an internet connection.
How AI is reshaping the market engine
Traditional algorithmic trading relied on rigid, deterministic rules: if stock A drops below its 50-day moving average while volume rises by 20%, sell. Fast, but brittle. Whenever the market regime shifts, static rules break down.
Modern AI systems work fundamentally differently. Rather than following a fixed checklist, they learn, adapt and combine numbers and language in real time:
- Sentiment and disclosure analysis: using natural language processing, AI models read earnings calls, central bank statements, regulatory filings and news in milliseconds. They pick up subtle shifts in tone or sentiment, often before those shifts show up in prices.
- Adaptive models: instead of static formulas, modern models continuously recalibrate based on what the market does, for example by adjusting stop-losses and position sizes when volatility rises.
- Emotionless execution: individual investors often lose money through fear of missing out or panic selling, not through a lack of ideas. Automated systems stick to predefined risk rules consistently, even in turbulent markets.
Breaking open the “quant club”
Until recently, building a quantitative strategy required fluency in programming languages such as Python or C++, access to expensive historical data and serious computing infrastructure. Today, new platforms largely remove that barrier: an idea written in plain language can be turned into executable trading logic.
| Dimension | Then: institutional | Now: widely accessible |
|---|---|---|
| Technical barrier | Specialist quantitative training and programming skills | Plain-language prompts and visual builders |
| Data access | Expensive proprietary feeds and manual analysis | AI summaries of filings, news and sentiment |
| Capital required | Large investments in data and infrastructure | An ordinary brokerage account |
| Strategy testing | Custom-built testing environments | Cloud-based historical simulations in seconds |
Academic research suggests that generative AI can narrow the information gap between individual investors and institutions, because it quickly summarises and dissects complex financial disclosures. In practice, three types of tools are emerging:
- No-code strategy builders, where you describe a trading idea in plain language, test it on years of historical data and have it executed automatically.
- AI signal platforms, which run large numbers of simulations across equities, ETFs and crypto to point out possible entry and exit moments.
- Integrated environments, which connect strategies directly to a broker, with automated order routing and risk management.
The new risks: when the machines agree
AI makes sophisticated techniques more accessible, but it does not remove risk. It changes its shape. When many market participants use similar models, new structural challenges arise:
- Herding and sudden volatility: market bodies and regulators warn that models trained on similar data can react to the same event at the same time. That can amplify short-term price swings and liquidity shortages.
- Overfitting to the past: a model can “memorise” historical noise rather than find a real predictive pattern. A backtest that looks flawless on past data can quickly fall apart in live trading when the market changes.
- Blind to the unexpected: models rely on historical patterns. In an unprecedented geopolitical shock or a sudden policy shift, they can misprice risk until people step in.
So the answer is not to trust a “black box” blindly. Even at professional firms, AI systems remain paired with human supervision and strict risk governance. The real edge lies in combining people and machines: AI to process disclosures, scan thousands of charts and take emotion out of execution, and human judgement for the wider context and the risk limits.
How we see it
At Vermeeren & Van Haute, automated trading is one of the six domains in which we invest for our own account. We do so with fixed rules, strict limits on position size and risk, and always under human supervision. Want to know more about what AI can and cannot do in financial markets today? Sign up for our info session on AI in financial markets.
This article contains general information and is not investment advice, an offer or a recommendation. Investing involves risk; you may lose (part of) your investment. Past results and historical simulations are no guarantee of future performance.