The entire financial circle is abuzz recently. Leading fund giants including E Fund, GF Fund, Harvest Fund, Southern Fund, China Merchants Fund and other major players have released their latest statements, with the concept of "working like a real researcher" going viral across the sector.
From Analysis to Inquiry, Fund Companies' "AI Researchers" Officially Take Up Their Posts
Recently, the scene of Tianhong Fund's "AI Researcher" attending the morning meeting and interacting with the investment research team has drawn widespread attention in the industry. "AI works like a real researcher" has become a portrayal of the current public offering industry re-examining business logic with AI.
In May 2025, the China Securities Regulatory Commission issued the Action Plan for Promoting the High-Quality Development of Publicly Offered Funds, encouraging fund companies to increase research and application of emerging technologies such as artificial intelligence and big data. A survey conducted by China Fund News shows that from reading research reports, generating meeting minutes, and calculating factors, to raising questions at key nodes now, the industry's focus on AI applications has shifted to how to use it more practically and deeply, and some leading institutions have entered a new cycle of large-scale application and platform-based supply.
"Working Like a Real Researcher"
At the morning meeting of Tianhong Fund's investment research team, the researcher's sharing was interrupted by a prompt on the shared screen: "Your profit forecast for the company's business module is significantly different from the consistent expectation in the brokerage research report. Please explain the reasons."
The one that sent the prompt is AI Challenge, which Tianhong's investment research personnel call the "AI Researcher". The "AI Researcher" will comprehensively interact with researchers based on sell-side, buy-side and internal past reports. Once the content told by the researcher is significantly different from the existing information, it will ask questions on the spot.
Some fund managers said that AI Challenge's opinions on some issues are "pointed and incisive", which can inspire the team to carry out further discussions.
It is introduced that AI Challenge is embedded in Tianhong Fund's investment research platform "Hongyan Zhice TIRD", which is a self-developed platform that is localized, private-domain, and deeply embedded in the investment research process. It aims to connect the three ends of research, investment and risk control, and has three core functions: investment research enabling tools, information collection and interaction tools, and management tools.
Many institutions predict that 2026 may be the "first year of investment research agents". At present, the industry tends to "re-examine business logic with AI", and some leading institutions have basically crossed the proof-of-concept stage and entered a new cycle of large-scale application and platform-based supply. From the perspective of deployment depth, the public offering industry has initially formed three echelons: E Fund, Harvest Fund, Southern Asset Management, Harvest Fund Management and other firms have built enterprise-level AI capability platforms, and AI penetration has been basically realized in the whole process of investment research; ICBC Credit Suisse Asset Management, China Merchants Fund, Guotai Asset Management, Fullgoal Fund and other firms have realized in-depth applications in scenarios such as pension operation, fixed income trading, and intelligent Q&A; other companies are still carrying out partial pilots.
Relevant personnel of Harvest Fund Management introduced that around the whole process of investment research, the company has built agent clusters for research, investment assistants, rule-based investment, and trading in key positions, forming a support chain from information acquisition, research report processing, viewpoint tracking to portfolio construction, transaction execution, attribution review and risk warning. As a task-driven position-based agent, the "AI Researcher" has taken over about 40% of the basic desk work of researchers - reading the internal knowledge base, learning the buy-side industry research framework from past reports, and then combining the latest market conditions, announcements and industry data to write research reports, give stock ratings, and form research dashboards, "working like a real researcher".
At the level of actual operation, fund companies particularly emphasize maintaining clear boundaries. E Fund uses the EWork platform as the carrier of intelligent applications, and builds functional modules such as investment research information retrieval and sorting, conversational Q&A, research analysis and material generation, data query and portfolio analysis assistance to assist researchers in improving efficiency, but does not replace manual judgment or investment decision-making.
Some fund companies have chosen a more "restrained" route. A certain fund company said that AI has been applied to the investment research process, including intelligent information processing, intelligent analysis, intelligent decision support and intelligent report generation, but it has not been anthropomorphized. Investment research personnel are still the main body outputting all conclusions, and AI only undertakes information preprocessing work.
AI Has Been Embedded in Multiple Business Scenarios
According to industry insiders, at present, AI has been deeply embedded in the complete end-to-end investment process of "research-investment-risk control", continuously producing verifiable expectation gaps, alternative signals and Alpha clues in long-term real account operation, contributing incremental returns to the investment portfolio, and has precipitated a variety of mature landing scenarios.
Xia Zheng'an, fund manager of Harvest Fund Management, introduced that the first category is dynamic identification of market narratives and main line opportunities. AI analyzes high-quality investment research corpora across the whole market in batches every day, structurally extracts massive scattered information, judges the crowding degree of different logics, distinguishes emerging main lines that are fully priced, gradually ebbing, and just sprouting, quickly sorts out the current market investment context, and outputs corresponding segmented tracks and target pools.
The second category is the mining of expectation gaps for low-attention targets. Relying on the quantitative screening framework, AI continuously scans targets in the market with low valuation positions, less sell-side coverage, and low trading heat, combines fundamental research data to measure the potential risk-return space of individual stocks, and pushes clues with attractive expected returns to the investment research end. Most of these targets are not within the mainstream hot spots of the market, and are easily ignored by traditional investment research frameworks, which are important sources of expectation gaps.
The third category is the iterative optimization of quantitative Alpha strategies. In the quantitative sector, based on the existing factor library, AI independently explores better solutions for factor combinations, weighting methods and strategy construction under reasonable constraints, continuously iterates the factor synthesis algorithm, optimizes indicators such as return stability and drawdown control of the strategy, and improves the overall risk-return characteristics. The optimized related strategies have been connected to real accounts, continuously providing stable incremental Alpha for the portfolio.
The fourth category is portfolio dynamic risk monitoring and early warning. The company has built a mature quantitative analysis system at the risk control end, covering multiple dimensions such as industry exposure, style exposure, benchmark deviation, and trading behavior analysis. Relying on this standardized framework, AI automatically completes multi-dimensional risk profiling and attribution analysis of the portfolio every day, timely identifies potential risk exposures and abnormal deviation signals and outputs early warning prompts, assists the team to troubleshoot hidden risks in advance, avoids unnecessary return loss caused by passive deviation, style mismatch, transaction friction, etc., and makes the risk-return structure of the portfolio more controllable.
It is introduced that China Merchants Fund's AI application has entered the stage of platform-based supply and replicable application from early technical pilots, embedded in core scenarios such as investment research, marketing, customer service, operation and risk control. In the field of investment research, applications such as intelligent meeting minutes, individual stock AI insight, research report translation and Q&A, AI credit evaluation report, intelligent financial report comment, and index constituent stock prediction have been implemented, and applications such as research viewpoint summary and review, public opinion aggregation and interest rate research assistance have been explored in the directions of equity, fixed income and ETF.
The core platform "Archimedes" integrated investment research system of GF Fund deeply serves core business lines such as fixed income, active equity, passive index, and cross-border investment, supporting the stable operation of large-scale assets. From the perspective of intelligence level, the company has realized landing applications in multiple core business lines such as intelligent investment research, intelligent risk control, intelligent trading, and intelligent attribution analysis.
It Will Reshape the "Mode of Production" of the Asset Management Industry
Industry insiders said that in the future, AI will reshape the "mode of production" of the asset management industry, but will not change the "production subject".
Relevant personnel of Harvest Fund Management believe that in the future, under the AI-native investment research system, it will be possible to quickly scan the whole market, cross-asset real-time dynamics, capture marginal changes in fundamentals, and make dynamic adjustments according to strategic constraints; AI will take over desk information processing, and humans will focus on business logic deduction, reverse thinking and portfolio construction. The information gap is being quickly smoothed out, and the source of excess returns will shift from "who can get public information faster and more comprehensively" to "who can read deeper logic from the same piece of information and convert it into portfolio expression", and the cognitive gap will become the core source of returns.
A large fund company based in Beijing believes that the popularization of artificial intelligence technology will reshape the production logic of public offering excess returns. However, what is changed is the mode of production, not the elimination of excess returns. The company emphasizes that AI has greatly lowered the threshold for information acquisition, and the advantage of information gap is gradually disappearing. The generation of excess returns is no longer limited to static research reports or viewpoints, but based on dynamic market changes and continuous research verification. Driven by AI, the investment research organization of fund companies evolves to a platform-based collaborative model. By building an intelligent knowledge middle platform, institutions convert personal experience into organizational assets, realize large-scale reuse of knowledge and cross-team collaboration, and improve the overall investment research strength.
For the future development trend of AI, relevant personnel of Harvest Fund Management said that first, improve the Agent basic platform, strengthen the task planning, arrangement and execution capabilities, and continuously improve the efficiency and quality of task execution; second, deepen the application of business Agents, deepen and refine the scenarios, cover various industry research scenarios and various asset investment scenarios, support personalized research demand configuration, and hand over the Skill construction and adjustment capabilities to business personnel; third, deeply integrate into the system process, embed AI capabilities into the integrated investment research platform, let Agents gradually take over the existing workflow, but set up manual review and decision-making links at key nodes to form a closed loop of "system + AI + human"; fourth, strengthen knowledge ontology governance, make good use of the company's internal structured and unstructured data, and incorporate high-quality AI output content into knowledge ontology governance.
For the challenges that may be faced in the future, a fund company in Shenzhen said that first, identify and falsify the output results of AI, identify hallucinations and pseudo signals; second, interpret industrial underlying cognition, management judgment, and non-standard soft information; third, conduct in-depth insight and forward-looking judgment on risk control constraints, portfolio architecture, and long-term fundamentals. In addition, it is necessary to strengthen the ability of independent thinking and reverse decision-making: in extreme market environments, AI may make misjudgments due to data limitations or model deviations. At this time, the independent thinking and reverse decision-making ability of investment research personnel is crucial.
Leading Fund Companies Settle in AI Workbenches to Seize the New Entrance of "Conversational" Services
"How many days does it take for fund redemption to arrive in the account?" "What impact will interest rate hikes have on the market?" In the past, investors had to repeatedly search between rule descriptions and financial news to figure out these questions. Now, these questions have a new destination: moving from the search box to the dialog box.
Recently, public offering funds have quietly set off a trend of "moving into" AI workbenches. Fund companies such as E Fund and Southern Asset Management have settled in Tencent WorkBuddy, and GF Fund has settled in ByteDance's "Doubao Work" platform. Professional fund knowledge is being packaged into AI conversation scenarios in batches.
When the slogan "where investors are, services go there" has become a technical reality, what fund companies have seized is not only the conversation interface, but also a new entrance to establish long-term trust relationships with hundreds of millions of investors.
Many Fund Companies Are "Moving into" AI Workbenches
The partnership between public offering funds and AI workbenches is becoming a new trend in the fund industry.
E Fund is the first fund company to settle in Tencent WorkBuddy. Its "E Fund Buddy Application" has three modes: "Understand Opportunities", "Select Funds", and "Hold and Accompany", which can reach 9 professional service scenarios, including hot spot tracking and interpretation in the pre-investment stage, fund manager viewpoints and theme concept analysis, comprehensive fund analysis, multi-dimensional screening and index investment zone in the mid-investment stage, and investor education learning, key time point companionship and full-touch service tour in the post-investment stage.
GF Fund was the first batch to settle in ByteDance's "Doubao Work" platform, and has launched the "GF Fund" connector and three major skills: "Fund Q&A Record" (answering transaction rules), "Viewpoint and Strategy Library" (interpreting market logic) and "Investment Research Translator" (disassembling research reports in plain language), connecting professional fund knowledge and market viewpoints into the AI conversation scenario. Relevant personnel of the company said that they hope to actively explore cooperation opportunities within the scope permitted by technical conditions and regulatory provisions, so that investors can obtain professional knowledge and market viewpoints more conveniently in familiar scenarios.
Southern Asset Management sets up five expert modules around the high-frequency demand of "how to select funds" on WorkBuddy: hot spot market analysis, fund selection, fund evaluation, index insight, and bond market observation, organizing the complex task of "selecting funds" that connects multiple pages into the same conversation.
From "Push Reading" to "On-Demand Call"
Translating institutional-level investment research content into words that ordinary investors can understand, and answering questions instantly in the form of conversation, this new model has brought profound changes to the industry's investor education and content operation methods.
Yang Anle, General Manager of the Customer Service and Sales Support Department of E Fund, said that this is an "upgrade" of the industry's investor education and content operation methods. In the past, investor education content was mainly distributed through channels such as traditional media and new media platforms, which belonged to "wide net casting" output. Now, with the help of AI, professional investment research and investor education content can be called on demand, combined with market data and user portraits, to achieve more accurate and personalized output.
In Yang Anle's view, this actually puts forward higher requirements for content operation: first, the underlying materials must be more rigorous and professional; second, the investor education content is more scenario-based and targeted; third, in addition to content, there must be distribution methods adapted to the AI platform.
"Conversational services broaden the scenarios of interaction with investors, which is an expansion and supplement to existing forms such as official accounts and investor education articles, rather than a replacement." Relevant personnel of GF Fund expressed another idea.
Interviewed institutions and third-party institutions judge that with the development of artificial intelligence, the way of investor education will undergo structural changes. In the past, investor education was to let investors "read articles". Now AI "calls materials to answer questions", forcing content production to move towards structuring, labeling and viewpoint orientation. TX Investment Consulting said that the production logic of investor education content has changed from "what the institution wants to talk about" to "what investors want to ask". This kind of conversational investor education can continuously accumulate investor portraits, changing from one-time content investor education to long-term companionship and relationship maintenance.
It is worth noting that the AI conversation entrance has already ushered in user migration, but there are still obstacles between "asking questions on the platform" and "completing fund decisions on the platform". What is the standard for judging that an entrance is worth occupying? Is it the volume of conversations, effective reach, or traceable conversion?
Relevant personnel of GF Fund said that from the perspective of investor service and companionship, the focus is on whether it can strengthen connections, effectively respond to investor questions, and enhance investors' understanding of funds.
Yang Anle's core consideration is whether it can provide more valuable services for more customers. Compared with short-term product scale and customer conversion, he pursues to establish a long-term trust relationship with customers.
Moving from "Product Center" to "Investor Center"
Will the actions of leading companies trigger a trend of fund companies settling in AI workbenches?
An industry insider in South China said that the settlement trend may have arrived, and the leading demonstration of leading institutions will drive other companies to follow. The new entrance amplifies the value of professional content and service capabilities, which will raise the access threshold. But in the short term, it is more of a supplementary reach, not a replacement. "It forms a complement to the traditional consignment platform - the AI workbench undertakes the responsibility of cognition and companionship, and the consignment channel undertakes transaction and conversion. The division of labor between the two cannot be mutually exclusive until the closed loop is opened."
TX Investment Consulting believes that other public offering institutions may follow up, and the AI workbench and the traditional consignment platform form a complementary relationship of "diversion - conversion".
From the long-term perspective of industry transformation, conversational AI services are expected to become an important technical starting point for the public offering industry to shift from "product center" to "investor center".
"The significance of conversational investor education services for the 'investor center' transformation does not lie in how many questions can be answered, but in the technical feasibility of realizing 'instant companionship when investors have needs'." TX Investment Consulting said that in the future, if the AI workbench can continuously reduce the marginal cost of "advisory" services and stably output high-quality and compliant companion responses to investors, it will become an important structural force driving the industry to transform to "investor center".
Some industry insiders said that real transformation requires institutions to keep up with the synchronization of assessment, product design and long-term companionship mechanisms, otherwise it will be difficult to bring substantial changes.
Liu Wei, General Manager of the Wealth Artificial Intelligence Department of E Fund, reminded that conversational AI financial services are easily understood by investors as "investment advisory", but it is essentially different from real investment advisory services. Investor education focuses on knowledge popularization and