On August 7, 2026, the investment research director of an asset management institution told Economic Observer reporters during the "2026 China Financial Market AI (Artificial Intelligence) Investment Research Summit" that investment research agents had independently built warehouses to buy up overseas crude oil futures and made profits.

On the day before attending the summit, the above-mentioned investment research director went to a third-party investment research institution for research and found that the latter's self-developed crude oil futures strategy intelligent agent had independently completed three Brent crude oil futures transactions in the past month without human intervention, all of which achieved profits.

On July 12th, the intelligent agent of the crude oil futures strategy opened a position at $80 per barrel and bought Brent crude oil futures higher. Two days later, the intelligent agent increased its long position in Brent crude oil futures near $83 per barrel. On July 23rd, when the Brent crude oil futures price hit $95 per barrel, the intelligent agent decided to liquidate all long positions and leave at a profit.

On the night of August 4th, the intelligent agent suddenly made another move and established a short position (buying down crude oil futures) when the Brent crude oil futures price was at $85 per barrel. One day later, when Brent crude oil futures fell to $79 per barrel, the agent quickly closed its short position.

The investment research director mentioned above revealed that in mid July, the intelligent agent continued to make predictions of a bullish outlook on Brent crude oil futures, while also refusing to chase higher on the grounds that "the expected return on opening a position above $90 is not attractive enough".

In his view, investment research intelligent agents now have the ability to think independently, and even have the ability to review and self optimize - they can judge whether the expected returns of each investment decision are reasonable.

This is a microcosm of the rise of investment research intelligent agents. The reporter learned that many financial institutions are working on developing investment research intelligent agents to comprehensively improve their investment research efficiency and capabilities.

Wang Pei, Chairman of Hongze Research, told the Economic Observer that currently, the reason why the financial market is gradually attaching importance to investment research intelligent agents is mainly based on three considerations: firstly, compared to financial market researchers who focus on dozens of data indicators that affect the correctness of investment decisions every day, intelligent agents can tirelessly "look through" hundreds of dimensions of data, capturing more potential investment opportunities and risk factors; The second is that the trading decisions and investment research perspective of the intelligent agent are "without any emotions", and can strictly implement the investment research framework and investment analysis logic set by financial institutions; The third is that the intelligent body will permanently remember all the frameworks and rules set by senior analysts, and there will be no "omissions or forgetfulness".

To some extent, investment research agents are replacing the most inefficient part of researchers' work - information data collection, information aggregation, and thinking and deduction under the fixed investment research process, allowing researchers to leave more time and energy to think about how to iteratively optimize their deep judgments, "said Wang Pei.

Tang Xiaodong, Co General Manager of Macro Strategy Department and General Manager of Quantitative Investment Department of Southern Fund, told Economic Observer that if financial investment research work is divided into financial research and financial investment, then financial research work can fully embrace AI. The reason is that intelligent agents can not only process the entire market target and million level feature factors in parallel, covering the entire market period and scanning potential opportunities and risks without blind spots, but also continuously output dynamic probabilities based on real-time data to make high-frequency adjustments to the conventional financial market situation, ensuring consistency and discipline in investment decisions.

Ran Qingkun, a senior solution architect at Alibaba Cloud, revealed that among financial institutions that have already implemented intelligent agents, 20% adopt a relatively conservative usage strategy, that is, they mainly use intelligent agents for multi-dimensional data collection and synthesis in the financial market, while strategy generation, risk management, daily backtracking, and review optimization are still manually completed; 50% of institutions go further by entrusting tasks such as data collection, data synthesis, and strategy generation to intelligent agents, while human labor mainly focuses on risk management and retrospective optimization; Another 30% of institutions simply entrust all tasks such as data collection, data synthesis, strategy generation, risk management, and review optimization to intelligent agents for independent completion.

At the summit, this also sparked heated discussions among financial institutions - will investment research intelligent agents fully replace traditional manual investment research processes?

The investment research director of a private equity fund revealed to reporters that in May this year, they will conduct internal testing and trial of their self-developed investment research intelligent agent in the A-share market. They found that over 30% of the investment decisions generated by the intelligent agent were difficult to 'explain', with most of them appearing to have a high winning rate, but their investment logic was completely 'black box'. This actually worries the investment research team that if the intelligent agent continues to make investment decisions independently, investment risks will eventually arise.

The investment manager of another asset management institution stated that they have developed an investment research risk control intelligent agent, which "absorbs" all the investment research risk control rules and disciplines of the asset management institution, and can intelligently verify and manage the risk control of all transaction orders placed by the asset management institution. However, the compliance department quickly raised objections to this, stating that if the intelligent agent encounters risk control loopholes and errors, resulting in investment losses, who will bear the corresponding responsibility. Under continuous pressure from the compliance department, this risk control intelligent agent has been discontinued.

Wang Pei stated that the essence of the operation of the investment research intelligent agent is its ability to efficiently and without omission execute every framework and rule set by the investment research team, but it does not define "which direction to pursue investment decision logic" for the team. Therefore, the depth of the framework and the choice of judgment are always in the hands of people.

Wang Pei believes that in the current era of increasing use of investment research intelligent agents, financial institutions should not focus on "big models with zero illusions", but lock in risks at key points. Intelligent agents cannot arbitrarily provide winning rates and odds for investment decisions, otherwise all risk control disciplines of financial institutions will become mere decorations

The human-machine collaboration capability will become indispensable in the field of investment research in the future, and this is also a new paradigm of division of labor in the era of investment research intelligent agents. The human brain is mainly responsible for formulating analysis directions and strengthening professional judgment on publicly available incremental information obtained from offline field research; the intelligent agent is responsible for strategy execution and improving investment research efficiency. The two are not interchangeable, but each plays its own role, allowing people to withdraw from repetitive work and return to places that truly require deep judgment, "said Wang Pei.

(Author Chen Zhi)