Volunteer Application Agent: Tencent is restrained, while Alibaba is aggressive

“Frankly speaking, we don't have competitors in the college entrance examination field. It's not that we don't pay attention to others. Over the years, from the first year to now, our entire team has only focused on what the users' needs are and how to turn technology into products.”

Zheng Shishou, the product leader of Alibaba's Qianwen Division, recently stated that the team has been developing college entrance examination tools for eight years, and AI is making a difference.

Before this year, the user perception for college entrance examination volunteer filling was Quark. However, after a year of integration and adjustment, the product that has come to the forefront is Qianwen. On June 10th, Qianwen released a full - cycle college entrance examination volunteer filling Agent. The key point in this official statement is Agent. Among the similar products of Tencent, Baidu, and Alibaba, the product features of Tencent's Yuanbao College Entrance Examination Assistant and Alibaba's Qianwen College Entrance Examination both focus on Agent. The real difference of Qianwen is that it has extended the product application cycle.

What supports Qianwen is the thinking, planning, execution, and reflection system of AI Agent, which also has the capabilities of active planning, long - term memory, and personalized services. Internally, it is believed that tool - based products cannot meet the long - cycle and personalized needs of college entrance examination volunteer filling, while Agent can cover the entire cycle of volunteer filling and even career planning.

The replacement of Quark College Entrance Examination by Qianwen College Entrance Examination and the iteration of other manufacturers' college entrance examination products reflect the replacement from search + tools to large models + Agent.

A Complex Math Problem

The college entrance examination volunteer filling product addresses a math problem related to matching: finding the optimal solution from the massive information of nearly 3,000 universities and over 2,000 majors across the country, with potentially hundreds of millions of combinations.

The standard answer may be found, but it is too slow and difficult for humans to solve this problem. Even with corresponding experts, the resources are extremely scarce, so tools are needed to fill the gap.

Before making a decision, candidates first need to master historical data and estimate the probability based on the previous years' admission rankings and score differences. When filling in the volunteers, they are also affected by multiple factors such as the school level, major strength, personal interests, location, and employment prospects.

Traditional tools have completed the pre - processing and standardization of information. Users can get corresponding answers as long as they search. However, in the real - world scenario, college entrance examination volunteer filling is a complex task chain that lasts for more than 20 days. Before the scores are released, candidates need to estimate their scores and initially determine their direction based on personal interests.

The most energy - consuming part is after the scores are released. Candidates need to match universities, majors, and regions according to their scores. After making a decision, they also need to carefully check to ensure there are no mistakes. What is the parallel volunteer system, which majors can be applied for with physics, chemistry, and biology, and what are the expectations of parents, etc. The real - world game far exceeds the first - principle of using tools and then leaving.

In the matter of volunteer filling, human resources are limited, and tools have their limitations, but this is exactly what AI is good at. In the combination space of volunteer filling, AI can estimate the value and eliminate obviously inferior combinations; predicting the admission probability is a natural application scenario for pattern recognition and regression analysis; and finally, recommending solutions for individuals can actually be covered by AI recommendation systems and multi - objective optimization.

To fill the huge gap left by tools, both Tencent and Alibaba have introduced Agent and provided two completely different solutions.

On June 5th, Tencent announced that Yuanbao and QQ Browser released the college entrance examination information Agent "Yuanbao College Entrance Examination Assistant", with the main function being the college entrance examination consultant Agent. Yuanbao is the mobile - end entrance, while QQ Browser provides the PC - end entrance. In the product usage process, candidates input their scores, get volunteer recommendations, screen universities and majors, and query previous years' data. Except for multi - round conversations and long - term memory, there isn't much difference from past products.

Tencent's relatively conservative product strategy is based on certain considerations. Volunteer filling is not a life - simulation game. One choice can have a very long - term impact. Enterprises can only provide auxiliary tools and cannot participate in the decision - making.

Five days later, Alibaba's Qianwen released its own college entrance examination volunteer filling Agent, which is much more aggressive than Tencent's. The product consists of three core capabilities: volunteer calendar, volunteer report, and volunteer Q&A. Qianwen Agent divides the college entrance examination volunteer filling scenario into three core scenarios, which are based on time, report reference, and dialogue fine - tuning, thus transforming volunteer filling from users' active search to an AI - assisted decision - making process.

After candidates fill in information such as their province, they enter an automated process. AI generates a schedule from the first day after the exam to the submission of volunteers, and pushes information on establishing cognition, positioning scores, exploring directions, pre - selecting solutions, and formal filling. It uses structured means to help candidates establish rational cognition.

In 2025, generative AI entered the field of volunteer filling for the first time, and all companies provided the ability to generate reports. However, due to the personalized nature of volunteer filling, it is difficult to simply evaluate what is good and what is bad. Therefore, the competition among products is not only about the accuracy of the reports but also the breadth of information provided.

Both Qianwen College Entrance Examination and Yuanbao College Entrance Examination Assistant provide recommendations and dynamic adjustments for "aggressive", "stable", and "safe" choices, but Qianwen sets more constraints for recommendations, such as major preferences and special qualifications.

In the past few years, volunteer Q&A has been the most obvious differentiating function among college entrance examination products. The five supports mentioned by Qianwen Agent at the conference can actually be summarized in plain language as dialogue memory, data accumulation, converting scores using the score - difference and ranking method, visualizing structured data, and changing past recommendations to active questioning.

The Decline of Tools is True, but Volunteer Filling Can Only be Assisted Driving

Before the era of large models + Agent, college entrance examination volunteer filling products were in a homogeneous competition.

Regardless of which giant company launched the product, in essence, they all output the information connection ability of the Internet, complete data accumulation, optimize algorithm testing, and finally implement it in the form of search. For example, Tencent first piloted the KNN + LM algorithm in 2017, then collected admission, university, and policy data, and launched the New College Entrance Examination Assistant in 2021.

Since 2021, college entrance examination volunteer filling, like New Year red envelopes, has gradually become a node for Internet companies to show their strength, convert traffic, and turn the data flywheel. During this period, college entrance examination tools have gone through the stages from search to volunteer filling, and then to generating volunteer reports last year. The product forms are all the same.

For example, Quark College Entrance Examination has gone through the stages of search query, launching the "aggressive, stable, and safe" tool, introducing the volunteer filling function, and adding the generation function last year. Continuously adding functions has not changed the foundation of search + tools. Moreover, for large companies, these tools do not have a moat. What is exclusive this year becomes infrastructure next year.

The product strategy of Qianwen College Entrance Examination Agent implies two business intentions.

Tools in homogeneous competition do not have a moat. Candidates don't really care who was the first to create the "aggressive, stable, and safe" concept. Querying the admission scores is no longer the evaluation standard for the quality of manufacturers' products but a basic requirement. The Internet has long eliminated the information gap in comparing universities and majors. In other words, search + tools have reached the end of diminishing marginal utility.

As long as it stays at the tool level, there is no moat for college entrance examination volunteer filling. The AI - generated volunteer analysis report launched last year has become a standard feature this year. Homogeneous competition has ultimately become an arms race at the functional level, and products cannot form differential pricing. Free services have become the norm.

The biggest variable brought by Agent is the reconstruction of the interaction itself. In the tool stage, users' behavior is active search, while Agent can turn the data flywheel by relying on the reasoning, memory, tool - calling of large models, and the continuous acquisition of user information.

Zheng Shishou, the product leader of Qianwen Division, said, "In the process of developing Qianwen, we want to let candidates and parents discover their own needs, hobbies, and interests more directly during the Q&A process. This is a different logic from the past, as it has changed from one - way use to anthropomorphic interaction."

Since the goals of candidates and parents are not clear, it is difficult for tools to intervene in subsequent actions. Qianwen's solution is to complete data alignment through multi - round conversations between humans and AI. There is nothing new about this solution: it imitates human behavior.

College entrance examination consultants not only have access to the college entrance examination data of various regions over the years, the enrollment situation of universities and majors, and rich experience, but more importantly, they can quickly understand the situation of candidates and parents. Whether it's the candidates or the parents who have the final say, whether they want to go to the coastal areas or the inland, and their job - seeking desires, etc.

Large models + Agent have access to structured and standardized data and can quickly use tools and make judgments. However, if candidates and parents are not motivated to speak up and provide information, the whole process cannot work.

However, strictly speaking, anthropomorphic interaction still does not have a moat. Qianwen Agent also has another more ambitious plan.

College entrance examination volunteer filling is a periodic scenario with a lack of long - term nature. After being Agent - enabled, through dialogue and constrained recommendations, it will expand to a wider range along the scope of college entrance examination and study, and complete the binding of user habits.

According to Guangzi Planet, this year's college entrance examination is a test. Qianwen will plan and develop subsequent products based on the test results. For example, it will update the calendar and Agent mechanism to provide a wider range of services. From this perspective, Qianwen College Entrance Examination Agent is just one of the many entry points for AI to penetrate into the learning and growth process.

Spend Tokens to Buy Trust?

Qianwen College Entrance Examination Agent does not deviate from Alibaba's AI strategy of "producing Tokens, transporting Tokens, and consuming Tokens".

College entrance examination volunteer filling provides a scenario with high consumption, high concurrency, and extremely high trust for Qianwen Agent. This provides momentum for increasing the stickiness and call volume of Qianwen's C - end users and driving Alibaba Cloud's MaaS platform.

In 2025, Quark first launched the AI volunteer report. Behind the 13 million reports generated, there was a huge investment in computing power. When Zheng Shishou was recently asked about the issue of computing power consumption, he did not give any direct response, only saying that "from the group's perspective, there are no restrictions on the whole thing, and full computing power support is provided."

The cost of Tokens is considerable, and it is a high - quality Token consumption. Every interaction between candidates and parents involves in - depth participation, real decision - making, and trust investment. Qianwen College Entrance Examination Agent consumes Tokens but gains trust assets in return.

In addition, the college entrance examination is also one of the national - level high - frequency and high - demand scenarios. Alibaba tries to enter every scenario and create a "Double 11" - like annual Token consumption event. By accumulating peak values, it aims to cover all aspects of life, learning, and work. The day after the release of Qianwen College Entrance Examination Agent, it also entered the field of the World Cup.

The scenario behind the college entrance examination is very deep, and the accumulation of trust assets can easily transform one - time consumption into long - term dependence. As for how far the World Cup prediction can go, apart from betting on football, there is no clear follow - up plan for now.

Maybe Qianwen just wants to take a "bet".

This article is from the WeChat official account “guangzi0088” (ID: TMTweb), written by Wu Xianzhi, and is published by 36Kr with authorization.