Tiger Brokers’ AI chatbot saves investors time but still comes with inaccuracies

“We’re focusing on using our TigerGPT to help users pick individual stocks … and go through a great quantity of financial news more efficiently,” Wang said.

TigerGPT’s represents one area in which companies are searching for practical uses of generative AI that can drive revenue, with industries such as financial services and pharmaceuticals among the early adopters.

Tiger Brokers’ bot has proven useful in reviewing real-time stock prices, exchange filings and analyst reports to generate quick answers to user queries about specific stocks, according to Wang. Compared with having people manually go through information on the trading platform, TigerGPT is able to cut hours of time, Wang said.

“Time matters in making a financial call; you need to know [the relevant information] in a short time,” Wang said.

However, the bot cannot yet replace human decision making, Wang said, as it may produce false information – an issue that plagues most generative AI products. Each answer from the chatbot is accompanied by a disclaimer. The company has been working on ways to improve the chatbot’s accuracy, according to the executive, including training it with sufficient data.

During initial testing of TigerGPT, the chatbot in some cases was generating inaccurate company financials, a situation that stemmed from a lack of historical training data. When Tiger discovered the issue, it started feeding the chatbot financial data going back 10 years, which improved results, according to the brokerage.

TigerGPT is available in markets including Singapore, Hong Kong, Macau, Australia and New Zealand, though most of its 70,000 users are based in Singapore and Hong Kong, where the company says it maintains contact with regulators to ensure compliance.

I