AI-Powered BIST Stock Analysis: How Multi-Model AI Commentary and Telegram Alarms Work

Reading technical indicators one by one and consistently interpreting the relationships between them is time-consuming. AI-powered analysis tools speed up this process by turning dozens of indicator values into a readable commentary and scenario. This article explains how these systems actually work, their limits, and how Lumina BIST implements this approach.

What Does AI-Assisted Technical Analysis Actually Do?

AI doesn't "predict" market data — it takes the numerical indicators you've computed (RSI, MACD, moving averages, volume data, etc.) and turns them into a natural-language commentary. Roughly, the process works like this:

1. The system gathers the current price data for your selected stock along with the calculated values of your selected indicators (potentially all 13).
2. This structured data is passed to a large language model (LLM) with an instruction like "interpret this technical data under the headings of overall trend, short-term risks, likely scenario, and a clear action suggestion."
3. The model produces a readable technical commentary based on the numerical data provided — but this is pattern recognition and language generation, not a "vision of the future" beyond the data itself.

There's a critical distinction here: AI is interpreting data, not predicting the future. Its output should always be evaluated with a "not investment advice" disclaimer.

Why Multiple Models Instead of Just One?

Large language models can sometimes emphasize the same data differently, or occasionally draw an incorrect or inconsistent inference (known as "hallucination"). Instead of blindly trusting a single model's commentary, comparing what multiple models (e.g. OpenAI GPT, Google Gemini, Anthropic Claude) say about the same dataset provides a form of cross-validation:

• If two models reach a similar conclusion, that's an additional signal of confidence that the interpretation isn't a random outlier.
• If two models reach different conclusions, that usually signals the market is genuinely giving mixed or uncertain signals — and that information itself is valuable.

Lumina BIST lets users pick a primary model during analysis and optionally view a second model's commentary alongside it (consensus mode).

How Do Telegram Alarms Work?

Watching price movements continuously isn't practical. Telegram alarms automate this need:

1. Set a target price: Create an alarm that triggers when a stock rises above or falls below a specific price.
2. Add an indicator condition (optional): Beyond the price condition, you can also define an indicator condition (e.g. "if RSI drops below 30"), combining the two with AND/OR logic.
3. Trigger: When the condition is met, the system automatically analyzes the current technical data and sends an instant Telegram notification with AI commentary explaining why the alarm fired.
4. Time-based alarms: You can also set up recurring alarms that send a technical analysis report at specific times of day, independent of any price condition.

AI Commentary at the Portfolio Level

Beyond single-stock analysis, you can also have AI evaluate the overall health of your entire portfolio. The system calculates each position's current gain/loss and its proximity to its stop-loss threshold, producing a commentary that summarizes which positions carry risk and the portfolio's overall outlook. This commentary doesn't give individual buy/sell advice — its purpose is general risk awareness.

Backtesting: Testing a Strategy Against Historical Data

Backtesting is used to see how an indicator combination would have performed historically. A specific entry/exit rule set (e.g. "buy when RSI drops below 30, sell when RSI rises above 70") is applied to real historical candle data, and metrics like win rate, average return, and maximum drawdown are calculated. This is a way to test a strategy's logic before risking live capital — but past performance is not a guarantee of future results.

Limits and Disclaimer

• AI commentary is a probability assessment based on technical data, not a certain prediction.
• Models may not reflect sudden news flow (KAP disclosures, macroeconomic shocks) in real time.
• No AI output replaces the personalized advice of a licensed investment advisor.
• This content and platform do not constitute investment advice; they provide technical data and educational information only.

Conclusion

AI-assisted analysis is a powerful tool for speeding up the process of interpreting technical indicators, and Telegram alarms automate that process further — but it's not a crystal ball. Its real value emerges when it's used as part of an informed decision-making process, alongside the fundamentals of technical analysis. On Lumina BIST you can compare multiple AI models, set price- and indicator-based Telegram alarms, get AI risk commentary for your portfolio, and test your strategies with backtesting.