“How is the gold price recently?” asked Zhang Chaoyang, founder of Sohu.com and a Ph.D. in Physics, at the launch event of Huatai Securities' "AI Zhang Le 1.0" version on January 26. Within five seconds, "Yomi," the voice assistant of AI Zhang Le, utilized its intelligent investment research Agent to dissect four core factors influencing the gold price: safe-haven demand, de-dollarization, interest rate cut expectations, and tight supply-demand dynamics, subsequently presenting the profit logic of various enterprises across the gold industry chain in the ensuing dialogue. As the securities industry's first AI-native trading application, AI Zhang Le, after 100 days of rapid iteration, resolutely chose to "evolve away" its menu, integrating high-frequency user functions into three core tools, adopting a posture perceived by outsiders as remarkably forward-thinking to address a central question in the current securities sector: how can general large language models evolve into professional and reliable investment assistants?
From "People Seeking Information" to "Information Finding People" "If you open AI Zhang Le for the first time, don't be surprised; this is an app without a menu," introduced Zhao Dayong, the product manager for AI Zhang Le. He explained that for ordinary users, using AI Zhang Le involves just three steps: first, see what deserves attention today; second, review what noteworthy information and signals exist for the stocks you follow; finally, delegate the execution of trades to AI Zhang Le. Traditional brokerage apps typically employ multi-level menus and tab structures, which often leads to high discovery costs for functions and lengthy response paths for investors. From its initial design phase, AI Zhang Le emphasized focusing on trading scenarios, eliminating the hide-and-seek style menus and initiating a task delegation mode to simplify the process of investors acquiring information and making investment decisions as much as possible. The newly launched 1.0 version of AI Zhang Le integrates these high-frequency user functions into three core tools: "Morning Briefing," responsible for organizing information and interpreting market directions; "Special Alerts," tasked with monitoring key market signals; and "Task Assistant," dedicated to efficiently completing execution actions, achieving a leap from "function-seeking" to "intent-driven" interaction.
"Morning Briefing" serves as the default homepage upon opening the app, providing personalized market interpretations before, during, and after market hours via an AI podcast format. By broadcasting only major events relevant to the user's holdings and watchlists, it helps investors filter out noise and conserve energy. Its "Rally Hunter" function focuses on capital flows, seeking opportunities for short-term traders, while the "Event Tracker" function delves into underlying variables affecting long-term corporate operations, serving medium to long-term investment strategies. "Special Alerts" acts as an "all-weather sentinel." When significant movements occur in a user's followed stocks—whether sudden fundamental changes or abnormal capital flows—the AI triggers instant warnings. During relatively calm market periods, it objectively screens for other high-value opportunities worth attention, helping investors avoid missing out or acting blindly. "Task Assistant" represents the most disruptive part of the new version, transforming complex investment processes into specific "AI tasks." From constructing stock selection logic to executing specific trades, users no longer need to navigate through secondary or even tertiary menus; instead, through simple swipes or voice commands, they can instruct the AI to perform complex backend calculations and matching.
A Financial Expert Standing on a Data Foundation "Yomi, I'm holding this stock at a loss; help me unwind the position." Following an on-site command, Yomi quickly displayed an unwinding tool, simultaneously providing backtesting data, profit-taking settings, and review cards. During a demonstration of the "Batch Profit-Taking" function, Yomi not only presented backtesting parameters but also transparently showed historical backtesting results. The technical team specifically emphasized that the AI's backtesting parameters are reference evaluations based on historical data, aimed at making investment decisions more rational rather than promising blind outcomes. Behind this seamless, "like moving one's own arm" smoothness lies an extremely complex "Multi-Expert Agent" technical architecture. According to Huatai's technical experts, AI Zhang Le does not simply call a single large language model API but has built a collaborative system where a main Agent coordinates multiple "Expert Agents." When a user presents a stock selection need, the main Agent quickly discerns the user's intent—whether to view charts, check financial reports, or perform technical analysis—then directs different expert Agents to call upon corresponding professional tools and databases, ensuring every AI response is evidence-based.
"We are not aiming to create the app with the most functions, but rather an AI investment assistant that is useful, easy to use, and trustworthy in key trading decision scenarios," said Wang Ling, head of AI Zhang Le and General Manager of Huatai Securities' Digital Operations Department. She added that future capability expansions will adhere to the principle of focusing on "trading scenarios," with everything aimed at improving users' decision-making efficiency and trading services; while on-screen indicator data is significantly reduced, more effective trading signals and events are highlighted. Zhang Chaoyang mentioned at the launch event that the AI Zhang Le development team did something very intuitive from a physics perspective: instead of training a base model from scratch, they stood on the shoulders of giants, conducting deep "Post-training." This is akin to sending a general scholar to a specialized finance class for intensive training, enabling causal chain analysis of market news by injecting real-time news and Huatai Securities' decades of accumulated private knowledge base. For instance, regarding "a drought occurring in a certain region," AI Zhang Le might deduce: drought → leads to decreased hydroelectric power generation → restricts production capacity of high energy-consumption metals (like electrolytic aluminum) → aluminum prices rise → benefits leading aluminum industry stocks. "AI Zhang Le integrates official exchange data and authoritative media information. Every response has clear data source annotations," Zhang Chaoyang particularly noted, stressing that in finance, "approximately accurate" equates to "completely wrong"; only data that has undergone strict admission screening can serve as a basis for investment, with "high-quality data" and "industry insider cognition" being the core advantages financial institutions possess when developing investment AI.
AI Bridging the Gap: Exploring Financial Technology Democratization Huatai Securities already has the Zhang Le APP with over 30 million users, so why create another AI Zhang Le? Wang Ling's response to this question was: "The issue of the power disparity between retail investors and institutions has long plagued the industry. Information gaps, technology gaps, and cognitive gaps objectively exist. AI Zhang Le is our bold attempt to use AI to bridge these divides in the investment field, allowing 200 million retail investors to access institutional financial services that previously cost hundreds of thousands." She explained that AI Zhang Le is committed to building proactive reasoning capabilities using graph-based narratives, using AI to explore the传导脉络 of events, and able to answer general investors with concise and clear logic: which enterprises are affected by a major news event, what the industry logic is, whether it impacts revenue, profit, or competitive landscape, and how capital markets might react. In this process, the Agent clearly displays the stock selection logic and each step, with this visualization effectively alleviating user concerns about "AI black-box decision-making," making the analysis traceable and understandable. Investors can also customize based on their own thinking, thereby building their own personalized Agents. Kong Xiang, Chief Non-Bank Financial Analyst at Guosen Securities, previously commented that the core of "AI Zhang Le" lies in deeply integrating financial technology into business operations and building core competitiveness through sustained, high-intensity investment. For the securities industry, competition in financial technology has shifted from "function stacking" to "ecosystem reconstruction," requiring the industry to focus on customers in the future and, under the premise of trust and security, achieve fundamental qualitative changes in services through technology. With the ongoing balance between regulation and innovation, AI technology is expected to further propel securities services towards smarter, more personalized, and more inclusive evolution.
Furthermore, security and compliance are the lifelines of the financial industry. Huatai Securities has also built multiple firewalls into AI Zhang Le, including sensitive word interception, query inspection, and real-time auditing of output content. Through a trusted whitelist mechanism and strong authentication mechanisms, they strive to ensure every data source called by the AI is authoritative and secure. As investor demand for intelligent trading continues to grow, many similar AI investment tools have emerged in the market. However, AI Zhang Le, by deeply focusing on the core scenario of "investment decision-making," has maintained a "transaction-centric" strategic positioning from the outset. Huatai Securities' decision to position AI Zhang Le as a standalone APP for differentiated strategy from the beginning also reflects the company's firm strategic judgment regarding the AI investment track. "AI cannot replace our thinking, but it can significantly shorten the time we need to acquire information and perform analysis and refinement. In the AI era, technological inclusivity has been achieved, breaking down information asymmetry and limitations in thinking capacity, allowing ordinary users to also access professionally valuable support. We also hope that in the AI era, all complex problems can become simple," Zhang Chaoyang concluded.
Comments