AI can genuinely be used to invest and even execute trades automatically. But there is a very important distinction between legitimate AI-assisted investing and the many advertisements that use “AI” as a marketing/scam word.
A real robo adviser, for example, can collect information about your goals, investment horizon and risk tolerance, then construct and manage a portfolio automatically.
1.0 How would AI actually trade for you?
Think of it as a pipeline:
1. Data collection The system continuously receives data such as:
- Share prices and trading volume
- Company financial statements
- Interest rates
- Economic indicators
- News
- Market sentiment
- Currency movements
- Sometimes alternative data such as satellite, web or social-media data
2. Analysis
Algorithms look for patterns and relationships.
For example, an AI might determine:
"Company A's earnings are improving, institutional buying is increasing, valuation is reasonable, and the overall sector is strengthening."
It could assign a probability or score rather than simply saying buy or sell.
3. Portfolio decision
The system then applies predetermined rules:
"Buy RM5,000 of A Reduce B by 20% Hold C Keep 15% cash"
A more sophisticated system may optimise the portfolio according to expected return and risk.
4. Actual trading
This is where genuine automation becomes interesting.
The AI/algorithm can connect to a broker through an API:
For example:
So, yes, a computer can literally buy and sell shares without you pressing the Buy/Sell button.
But here's the catch
AI doesn't possess a crystal ball.
Suppose an AI analyses 10,000 pieces of information and concludes:
"There is a 68% probability that this stock will rise."
That still means there is a 32% probability that its prediction is wrong.
And markets are particularly difficult because once many traders discover the same pattern, the advantage can disappear.
There are also problems such as:
- Bad or incomplete data
- Sudden geopolitical events
- Market crashes
- Overfitting historical data
- AI hallucinations or erroneous reasoning
- Trading costs and slippage
- Excessive trading
- Model failure
- Unexpected correlations
Recent research is also highlighting an important issue: an AI recommendation isn't necessarily a valid executable trading decision once fees, portfolio constraints and other real-world limitations are considered.
And this is where I would be VERY careful with the advertisements you're seeing.
There is a huge difference between:
"We provide AI-assisted portfolio management through a regulated investment platform."
and
"Our AI trading system makes 10% every week automatically with zero risk."
The second should set off every alarm bell.
The SEC, FINRA and NASAA have specifically warned about fraudulent investment schemes using AI claims, including promises that proprietary AI systems "can't lose" or can guarantee exceptional returns.
And this isn't merely a US problem.
In Malaysia, police investigated a 2026 case where a victim lost RM914,209 after being persuaded to invest in an alleged AI-powered automated trading system. The victim was shown apparent profits but subsequently couldn't withdraw the money.
Another Malaysian case involved an alleged AI investment platform promising 20–30% returns within 24 hours; the victim reportedly lost RM343,190.
The Securities Commission Malaysia has also warned about AI-generated deepfake videos impersonating prominent personalities and reputable companies to promote fraudulent investments.
And there is an interesting fourth category:
AI as your investment research assistant.
Instead of giving your money to an unknown "AI trading company", you could use AI to help you:
That's already quite different from an advertisement saying:
"AI finds the next winning stock."
The system is actually managing risk and allocation, not simply trying to predict tomorrow's winner.
6. Now the AI receives new information
Suppose Stock A has:
Rising earnings, Increasing trading volume, Positive sector outlook, Improving margins, Reasonable valuation and Positive news sentiment
The AI might produce something like:
Stock A
- Expected return: +8.2%,
- Probability of positive return: 67%,
- Risk score = Moderate,
- Signal: BUY
But that doesn't mean: BUY RM10,000!
The next layer says:
Maximum allocation = 20%, Therefore: BUY RM2,000
7. Then something unexpected happens
Imagine the next morning a major announcement causes Stock A to fall 7%.
The AI may recognise:
New information has invalidated part of the original investment thesis.
The system could then decide:
HOLD, REDUCE, or SELL.
For example:
- Current position: RM2,000,
- Sell 50% = RM1,000
- Remaining exposure = RM1,000
That is algorithmic risk management.
8. Now let's make it more sophisticated
A genuine system could have several models working together.
Model 1 : Fundamental analysis
It examines: Revenue, Profit, Cash flow, Debt, ROE, Margins, Valuation, Earnings growth
Model 2 : Technical analysis
It examines: Moving averages, Momentum, Volume, Volatility, Price trends
Model 3 : Sentiment analysis
It analyses: Financial news, Company announcements, Analyst reports, Possibly social-media sentiment
Model 4 : Macro analysis
It monitors: Interest rates, Inflation, Currency, Commodity prices, Economic growth
Then the system combines them.
For example:
- Fundamental score 82/100
- Technical score 71/100
- Sentiment score 76/100
- Macro score 61/100
- Risk score 35/100
- Overall score 73/100
- Decision: BUY
- Maximum allocation: 15%
That's much closer to what an institutional quantitative system might look like.
9. But here's the REALLY important part
Suppose the AI says: BUY, the system still shouldn't blindly trade.
It should ask: "How much?"
That's called position sizing.
For example:
- Portfolio = RM10,000
- Maximum position = 15%
- Maximum risk per trade = 1%
So perhaps it buys only RM1,500. This is one of the things that many flashy AI-investment advertisements conveniently don't explain.
10. And then there's the broker
This is where actual automation occurs. A legitimate broker can provide an API. The software sends something like:
- BUY
- Ticker: ABC
- Quantity: 500 shares
- Order type: LIMIT
- Maximum price: RM3.00
The broker receives the instruction.
If the conditions are satisfied: Broker → Bursa Malaysia → Order matched
The portfolio software then receives confirmation:
- ORDER EXECUTED
- ABC
- 500 shares
- RM2.98
Value: RM1,490
The AI system updates your portfolio automatically.
11. It can even operate 24/7 but not necessarily trade 24/7
The system can continuously monitor information.
For example:
- 08:00 Market preparation
- 09:00 Market opens
- 09:30 Scan prices
- 10:00 Scan news
- 11:00 Recalculate signals
- 12:00 Risk assessment
- 14:00 Portfolio optimisation
- 15:30 Rebalance if required
- 17:00 Generate daily report
- 17:00 Generate daily report
It doesn't necessarily buy and sell constantly. In fact, excessive trading can be a major problem.
12. Now imagine the AI gets it wrong
This is the part advertisements rarely show.
Suppose:
- AI predicts 75% probability of a rise, It buys RM1,500.
Then a completely unexpected event occurs. The stock falls 15%.
You lose approximately: RM225
The AI doesn't say: "Sorry, I'm an AI."
It simply has a losing trade. And this can happen repeatedly.
That's why risk management is arguably more important than the AI prediction itself.
13. MOST DANGEROUS VERSION
Imagine someone tells you: "Deposit RM10,000 into our AI platform."
Then: "Our AI trades automatically."
Then you see: RM10,000 → RM12,400
You become excited.
Then: RM12,400 → RM16,700
They tell you: "Our AI is extremely successful."
You deposit another RM30,000.
The dashboard shows: RM46,700
But when you try to withdraw:
- "Pay tax first." Then:
- "Pay verification fee." Then:
- "Upgrade your AI account."
Eventually you discover that the "profit" displayed on the website was never actually sitting in a brokerage account belonging to you.
That isn't AI investing but a scam wearing an AI costume.
14. If I were evaluating an "AI investment" advertisement
I'd ask these questions before putting RM1 into it:
- Who is the licensed broker?
- Where is my money actually held?
- Can I see the brokerage account myself?
- Who regulates the company?
- Can I withdraw directly without paying an additional "tax", "unlock fee" or "AI upgrade"?
- What is the audited performance over 3–5 years?
- What happens during a 30% market crash?
- What are the trading fees?
- What is the maximum historical drawdown?
- Can they explain the investment strategy without saying "our proprietary AI algorithm"?
That last one is particularly useful.
"Proprietary" doesn't automatically mean sophisticated.
The fascinating part
You don't actually need an autonomous AI trading bot to benefit from AI.
A much safer architecture for an individual investor could be:
You, Your AI research, Your investment decision, Regulated broker, Your money
rather than:
You, unknown AI company, unknown algorithm, unknown broker, goodbye RM50,000
And if you wanted to experiment, you could even build the first version without risking real money:
Historical data, AI strategy, simulated trades, paper portfolio, measure performance, only then consider real money.
That is called backtesting/paper trading, and it is where I would start.
Good Luck
