The author is a senior market analyst at amana and an accredited investment manager under CMA UAE
Spend a few days searching for investment advice online and sooner or later you will be offered an AI trading bot. The bot watches the markets around the clock and trades without the fear, greed or second-guessing that gets human investors into trouble. Some sellers claim win rates of 80 or even 90 per cent. You fund an account and the bot gets to work while you sleep.
There is one thing I have never quite understood about this business. If somebody has really developed software capable of producing exceptional returns with such consistency, why sell it to strangers for a monthly subscription?
Banks, hedge funds and proprietary trading firms have automated their trading for decades, executing orders at speeds no human could match. Virtu Financial, the American electronic trading firm, famously disclosed just one losing day in its first six years. Extraordinary, but it tells us little about the retail bots now advertised online.
Part of the confusion is that almost every form of computer-assisted trading is now called AI. Quantitative and high-frequency firms spend fortunes on researchers, data and infrastructure. A few-hundred-dollar retail subscription is unlikely to buy anything comparable.
There are plenty of sensible uses for AI in investing. It can examine financial statements, compare companies and spot patterns an analyst might take days to find. It can also curb familiar behavioural mistakes: a machine has no sentimental attachment to the price at which you bought a stock, although whether its strategy is any better than yours is another matter.
The difficulty begins when investors surrender control to a system they may not understand, particularly when the only evidence of its success comes from its seller.
A 2024 Review of Finance study examined experimental markets where people could trade manually and with robots. These were conventional execution algorithms rather than today’s autonomous AI. Participants who combined automated and manual trading tended to fare better, while greater reliance on robots was associated with lower earnings.
A 2026 experiment in the Journal of Economic Behavior & Organization introduced an autonomous AI trader into markets populated by humans. The machine topped most sessions. Yet its presence did not significantly improve pricing efficiency, while the human participants grew poorer.
Both were controlled experiments, but they raise a question. If increasingly sophisticated machines compete for the same opportunities, how long will those opportunities last? Markets tend to erode advantages once enough traders find them.
Back-tested returns present another difficulty. A model tested against years of data can post an excellent record, but if that same data was used repeatedly to refine it, the model may simply have mastered conditions that have already occurred.
The real test comes when markets change. A strategy built during years of low interest rates may struggle when borrowing costs rise sharply. Correlations shift, liquidity disappears and unexpected events produce movements historical models cannot anticipate. Human traders are hardly immune to this either. What matters is whether a system can limit losses when its assumptions stop working.
Transaction costs, spreads and slippage eat into profits, while leverage can turn a small market movement into a substantial loss. What shines in a simulation can disappoint once real orders are placed.
Britain’s Financial Conduct Authority has warned that approximately 80 per cent of customers lose money trading contracts for difference, or CFDs. That figure covers CFD trading generally rather than AI bots, but these are the customers to whom many automated products are marketed.
The more extravagant claims are hard to assess because independently verified, long-term live records rarely appear alongside the advertisements. Even a genuine 90 per cent win rate means little. A bot could win nine trades worth $100 each and lose $2,000 on the tenth. Its success rate would remain 90 per cent, although the investor would be down $1,100 before costs.
Actual returns, maximum drawdown, leverage and verified trading history matter far more than the percentage of winning trades. So do the company behind the platform, custody of client funds and regulatory status.
The business model deserves scrutiny too. A company selling trading software may earn from subscriptions, commissions or referrals, whether or not its customers profit. Its interests may be quite different from theirs.
Professional firms use automation extensively, but surround it with risk limits, monitoring and procedures for intervention. Handing an opaque retail product control of an account on the strength of a few screenshots is a very different proposition.
The financial industry has been selling effortless trading for years. Forex robots, crypto bots, automated signals and copy-trading platforms all promised profits without understanding the markets. AI is a far more powerful technology, but it has also given an old sales pitch a fashionable new vocabulary.
AI will undoubtedly play a growing role in investing, especially in research, analysis and execution. Whether an affordable autonomous bot can consistently deliver exceptional returns unsupervised is far less certain.
If someone offers me a bot that supposedly wins nine trades out of ten while I sleep, I would certainly like to see the independently verified results. I would also like to know why they need my subscription money.
Disclaimer: This article is intended for informational and educational purposes only and does not constitute investment advice. All investments carry risk. Past performance is not indicative of future result
Source: Khaleej Times