A public fund manager stated that while professionals will not be replaced by AI, they risk being overtaken by those adept at using AI, especially in the field of fund investment research, which blends science with art and rationality with intuition. Recently, a wave of enthusiasm for "raising AI agents," akin to cultivating lobsters, has spread from the tech sector to the heart of finance. AI Agents, represented by OpenClaw, are increasingly capturing the attention of public funds.

It is understood that multiple fund companies are currently cautiously evaluating the application of this tool in fund investment research. Some fund managers, particularly quantitative managers, have begun experimenting with OpenClaw for strategy development. AI is gradually evolving from a "super tool" into an "autonomous collaborator." However, on the flip side, the fund industry is also re-examining the impact of AI on traditional investment research models. Whether in processing vast amounts of financial data, identifying signals in quantitative investing, or even in previously complex research models, large language models and AI Agents are subtly influencing fund research. The public fund industry is undergoing a gentle yet profound "workflow revolution," but it simultaneously faces threats such as human-machine substitution and data leakage.

The "raising lobsters" trend has heated up the fund circle. A quantitative fund manager in Shanghai recently described how OpenClaw assists his work, noting that initial expectations were for an intern-like helper for backtesting scripts and data processing. However, after two weeks of use, he found its autonomy strong, capable of independently extracting valuable factors from raw data around the clock, broadening sources of Alpha with high accuracy—like having a senior assistant available 24/7.

Recently, open-source AI Agent projects like OpenClaw have become popular, sparking a "lobster-raising" craze across society, including the information-intensive and decision-complex field of fund investment research. BoShi Fund's Chief Digital Officer, Che Hongyuan, revealed that teams within BoShi are already using OpenClaw on public clouds under compliance premises, while also exploring scenarios for secure internal use of domestic software. Similarly, YiFangDa Fund has formed a specialized team to conduct functional verification and technical exploration of OpenClaw in an isolated network environment, though it has not yet progressed to production deployment. According to a representative from YiFangDa's fintech division, application scenarios focus on tasks like automated market information collection and analysis, and corporate data governance.

Che Hongyuan expressed that OpenClaw, as an open-source, deeply customizable AI agent, has ignited a new wave of enthusiasm for AI applications in the public fund industry with its "AI execution" capability, a significance that far exceeds the tool itself. He described OpenClaw as primarily an individual-focused agent that could greatly unleash personal innovation. In application, individual initiative is crucial; currently, investment researchers are the early adopters, as it provides them with a "super digital assistant" that helps release individual creativity and productivity.

YiMin Fund believes that for fund investment research, OpenClaw is not merely an enhancement of existing tools but is gradually triggering a gentle yet profound workflow revolution. Traditional research tools like Excel, Wind, and Bloomberg are fundamentally "passive responders"—humans input instructions, and tools output results. In contrast, the core breakthrough of AI Agents like OpenClaw is "active execution," enabling them to autonomously complete closed loops from information retrieval and data organization to preliminary analysis and result feedback based on preset goals. For instance, tasks that previously required researchers to spend one to two days consolidating public sentiment or financial statement data for a specific sector can now be automated 24/7 through configured OpenClaw skill modules. Researchers can then focus on data interpretation and logic validation, essentially restructuring the research workflow rather than simply improving efficiency.

Wang Ying, Deputy Director of Quantitative Investment at CITIC-Prudential Fund, stated that their quantitative team has long integrated AI technology into daily research systems. Currently, quantitatively factors trained via machine learning account for about 30% of their strategies, primarily applied in price-volume-based trading strategies. She explained that trading signals identified by AI models tend to yield better returns when executed on the same day, as AI excels at capturing short-term, liquidity-driven price surges. Intervening at these moments not only seizes fleeting opportunities but also reduces transaction costs due to ample liquidity, with the entire process enabling automated signal generation.

From large language models that serve as "super brains" to AI Agents capable of autonomous planning and execution, the rapid evolution of AI is directly impacting foundational, repetitive tasks in traditional investment research, such as information gathering, data organization, and report writing. The question for public fund professionals, especially those in investment research, is whether they will face displacement akin to "textile workers" during the first Industrial Revolution.

A public fund manager in Southern China compared AI to a "mature intern" or "research novice," proficient in tasks like data collection, organization, cross-verification, and even simple analysis. By handling these foundational tasks, AI frees up researchers to focus on areas where AI currently falls short. Wang Yue, a fund manager at MinSheng Jiayin Fund, also believes that the skill sets of AI and humans in investment research do not overlap but are complementary. A good researcher continuously poses insightful questions to identify core variables in an industry or company, while a good AI provides sharp, accurate answers to those questions, enhancing research efficiency without possessing true reasoning or thinking capabilities.

Wei Yu, a fund manager at HSBC Jintrust Fund, agrees that AI cannot currently replace fund managers or researchers. It serves more as a research assistant, processing historical data to identify key points and patterns. Researchers and managers then use their accumulated cognitive abilities to make more accurate industry judgments and investment decisions based on this information.

The Southern China fund manager added that many tasks remain beyond AI's reach, such as on-site due diligence. Interacting with company management in person provides crucial insights into their working state—a seemingly intuitive factor that often foreshadows company performance. Similarly, analyzing non-public information under compliance, which holds significant analytical value, is something AI cannot do, as it only processes existing materials.

While AI has become a powerful tool in investment research, the professional expertise of fund managers and entire research teams remains distinct and even more critical. YiMin Fund posits that AI replaces low-value-added tasks within a role, not the role itself, threatening individuals unwilling to adapt or with limited skills, not core research professionals.

In the AI era, the "information gap" in markets will narrow, as AI can quickly gather and analyze vast amounts of data, providing nearly all institutions with the same foundational information. Therefore, future excess returns will stem not from who accesses information faster, but from who can interpret information more deeply, judge trends more accurately, and manage risks more effectively. Simply put, competition shifts from computational power to algorithms, which form the core of a fund manager's personal Alpha and a fund company's research edge.

Wang Yue emphasized that research depth is more valued than information breadth. Researchers must ask the critical questions rather than seek all-encompassing answers. Infinite information exists; identifying and敏锐地 grasping the core variable is the source of a researcher's excess returns.

Che Hongyuan added that by automating the processing of multi-modal information, AI enhances efficiency, compelling researchers to transition towards higher-order skills like deep logical deduction, industrial insight, and cross-verification. This evolution promises to advance research systems towards a "human-machine collaborative" network structure, moving away from the linear "researcher recommends, manager decides" model. Researchers may soon collaborate with AI on clue discovery, strategy building, and risk control.

The Southern China fund manager candidly stated, "We won't be replaced by AI, but we will be replaced by those proficient in using AI, especially in a field like investment research that intertwines science and art, rationality and intuition." From a corporate perspective, building a research edge involves creating an ecosystem suited to the company, requiring sound mechanisms, culture, talent, and tools—inevitably AI-related in the future. Ultimately, within this environment, individuals leverage their strengths while drawing nourishment from a unified information platform, forming a coordinated, cohesive ecosystem.

Embracing efficiency necessitates heightened vigilance against risks. While AI undoubtedly makes daily research more efficient, many public funds recognize it as a double-edged sword—it significantly boosts research efficiency but harbors numerous risks that, if unaddressed, could lead to investment losses.

YiMin Fund's Quantitative Finance Lab highlighted the need to警惕 the "black box risk" of AI models as the most critical concern. Currently, most AI models, especially deep learning models, operate with "unexplainable" logic—inputs and outputs are known, but the reasoning process is not. This risk manifests in two ways: First, "pseudo-effective" factor mining, where AI may identify seemingly significant factors that are merely artifacts of historical data fitting, failing in future markets or causing losses. Second, "misleading" decision suggestions, where AI might propose seemingly reasonable advice based on flawed logic or biased data. Blind reliance by fund managers could lead to erroneous investment decisions. For example, AI might recommend buying a stock based on strong historical performance, overlooking deteriorated fundamentals, potentially causing significant losses. Furthermore, the unexplainable nature of AI models makes risk tracing and root cause analysis difficult.

Wang Ying also maintains deep caution regarding the comprehensive application of AI. From a price-volume trading strategy perspective, she pointed out that markets are "adaptive." Trading behavior in the A-share market is a dynamically self-adjusting process. Historical data used for training inherently contains past market participant behaviors, and once a model begins trading, its actions become new market data, influencing the market itself—essentially, a model continuously affecting its own environment. This feedback loop causes the excess returns of AI-trained factors, particularly price-volume factors, to exhibit high volatility. She cited examples where machine learning factors performed excellently in 2023 but showed巨大 volatility recently, illustrating that the source of profits can also be the source of future losses. The greatest challenge, she admitted, is knowing when to deactivate such a model.

Wang Yue added that in fund investment research applications, special caution is needed against sensitive information leakage, a core concern with most current AI systems. Therefore, interactions with AI primarily involve speculative dialogue and processing public information, underpinned by robust privacy protection mechanisms and measures to prevent AI from accessing confidential company operational data.