Big tech giants fight tooth and nail at WAIC: Alibaba and Tencent go head-to-head, the full-scale battle for seizing the intelligent agent entrance has kicked off across the board.

After the "Hundred Model Battle" cools down, a new round of AI portal wars among major tech giants is intensifying, with this phase centered on the competition for agent portals.

On July 21, according to a report by *Caijing*, Alibaba is set to launch Qwen Office, which has already integrated three Agent products: QoderWork, Wukong, and MuleRun. Qwen Office will be Alibaba's flagship product targeting the Agent office market. Meanwhile, a reporter from *Time Weekly* learned from an informed source that Tencent has issued an internal notice to integrate the Workbuddy and QClaw teams.

The head-to-head competition between the two giants was also prominently on display at the 2026 World Artificial Intelligence Conference and the Global Summit on AI Governance (hereinafter referred to as "WAIC 2026"), which concluded on July 20. Over the four-day event, more than 1,100 exhibiting companies showcased over 4,400 technological products. Leading tech giants at the forefront of the industry also engaged in direct rivalry during this conference.

At Hall H1 of the Shanghai World Expo Exhibition & Convention Center, the exhibition booths of Alibaba and Tencent face each other directly.

Under the theme "Hey, My AI Buddy", Tencent showcased the evolution of AI into a productivity and life-assisting capability across various fields, ranging from the recently popular Workbuddy to Qclaw and ima for individual users, as well as the debut of the Hunyuan Hy3 large model. At Alibaba's booth, a product named "Qwen Office" was already on display. On-site staff told the *Time Weekly* reporter that the product is currently in internal testing and is operated by the DingTalk team.

Beyond desktop agents, Alibaba and Tencent have also launched direct competition in smart hardware terminals. During WAIC, Tencent Cloud WorkBuddy and Li Weike Technology announced a strategic ecological partnership and released the first X-AI memory glasses integrated with WorkBuddy. In the same period, Tencent's self-developed efficiency agent WorkBuddy officially launched its standalone app, with iOS, Android, and Hongmeng versions going live simultaneously.

During this year's WAIC, Qwen Glasses directly upgraded its AI glasses to agent glasses. After the upgrade, the glasses can enhance service and decision-making capabilities through agents, and call third-party Skills and Agents as needed. At the same time, Qwen's first AI agent headset made its official debut.

At this year's WAIC site, Ant Group focused on showcasing three major achievements during the event: the latest application implementations in AI health, AI life and other fields, including A Fu and A Bao; the agent business ecosystem layout centered on payment; and continuous investment in technical foundations, including Ant Lingbo general-purpose robot brain, OceanBase AI database, Ant Privy Computing HOP3.0, and the native secure agent operating system Agent OS, to provide underlying support for AI development. Baidu brought a full suite of agents including Baidu Buddy, Baidu One Shot, General Agent GenFlow4.0, Baidu Miaoda 3.5 version, and Baidu Famou 2.0.

A report from the International Data Corporation (IDC) also confirms the current industry changes. The industry's first "DAA Research Report" shared by IDC shows that the number of global active Agents reached 28.6 million in 2025; it is expected to hit 79.4 million in 2026 and grow to 2.216 billion by 2030.

As Yin Qi, Chairman of Stepfun and Chairman of Qianli Technology, judged at WAIC: The AI industry evolves in three stages. The first stage is pre-trained large models + conversational applications, the second stage is reinforcement learning reasoning + code development assistance, and the third stage is industrial transformation brought by embodied agents, whose influence will exceed the sum of the transformations of the PC internet and mobile internet. And general-purpose agents for physical scenarios are the core mainline of the third wave.

The wave of artificial intelligence is surging forward, and a new round of agent portal wars is being ignited. In this major technological transformation, the top priority for major tech enterprises is to hold their ground and stay in the game, so as to maintain their competitiveness continuously.

Alibaba and Tencent Go Head-to-Head in the Agent Arena

Both Alibaba and Tencent have taken competition paths that are nearly identical in essence: evolving from simple question-and-answer interactions to autonomous execution, which ultimately boils down to a battle for portals.

Tencent's WorkBuddy booth attracted a large number of visitors for hands-on experience. This full-scenario workplace AI agent workbench launched in March this year can help users complete work and deliver results with just a simple sentence. A *Time Weekly* reporter experienced WorkBuddy on-site, issuing the instruction "Sort out the visiting guide for the 2026 WAIC conference". Soon, WorkBuddy not only provided a one-sentence conclusion and five core highlights, but also synchronously delivered the "WAIC2026 Agent Special Report".

According to Tencent, the core concept of this agent is to go from "one sentence" to "delivery" — it is no longer limited to conversations, but can independently break down tasks, plan processes, call tools, complete work, and deliver results.

"At present, AI is transforming from a tool to a partner, from 'capable of generating content' to 'capable of completing tasks', and from a single-point capability to a systematic ecosystem," said Li Kan, Vice President of Tencent Group and General Manager of the East China Headquarters.

At the Alibaba booth opposite, a desktop agent labeled "Qwen Office" also attracted many experience users. By scanning the QR code on site to leave information, users can get priority access to Qwen Office usage rights. The *Time Weekly* reporter found through on-site experience that this agent has core functions similar to Tencent's WorkBuddy: users can either manually select models on their own, or let the system intelligently allocate models. On-site staff told the reporter that the product is currently in internal testing.

Their paths are similar: WorkBuddy is also being embedded into corresponding smart hardware terminals. The X-AI memory glasses are equipped with the self-developed WakeeMemory OS, which can continuously perceive real work scenarios, and the organized information will be automatically synchronized to WorkBuddy. That is to say, based on the long-term accumulated work memory, WakeeMemory can continuously understand users' work habits, task priorities, and cooperative relationships, organize fragmented information into action items with complete context, and then hand them over to WorkBuddy for subsequent execution.

This means that what AI receives is no longer a loose and isolated single-point instruction, but a complete workflow containing background, character relationships, time nodes, and priorities. Different from traditional AI office tools that mainly rely on active user input, the Li Weike X-AI memory glasses serve as the portal for WorkBuddy to connect to the real world. WorkBuddy is also accelerating the expansion of this unobtrusive, intelligent experience to more daily scenarios.

The logic behind upgrading Qwen AI glasses to agent glasses is similar. According to relevant personnel from Qwen, in addition to service and decision-making capabilities, Qwen has also launched a series of new hardware technologies such as full-duplex voice, eye-tracking, and vital sign monitoring to enhance the perception and interaction capabilities of agent glasses for the physical world. These hardware technologies will be applied to Qwen AI agent glasses within this year; the already released Qwen AI Glasses S1 and G1 will also introduce Agent capabilities through OTA to further improve personalized task understanding and execution capabilities.

Wu Jianjun, General Manager of Qwen AI Hardware Products, stated at WAIC: The starting point of interaction in the AI hardware era lies in the agent's complete understanding of users and the environment. "An agent without hardware cannot reach the real world; hardware without agents is just a cold tool. But when both are perfected to the extreme, they will become the only optimal solution. We develop the best hardware and combine it with the best agents to create the best AI assistant for users."

Is Autonomous Evolution the Essence Behind Agent Competition?

The year 2026 is regarded by the industry as the first year of AI application implementation. Richard Sutton, the 2024 Turing Award winner and a professor of computer science at the University of Alberta known as the "Father of Reinforcement Learning", believes that the core inflection point of AI development is the shift from static data to experience-driven progress.

At this year's WAIC, the consensus demonstrated by the competition among major tech giants is that with the continuous improvement of model capabilities, relying solely on scale expansion is no longer sufficient to meet the needs of making AI truly useful to humans. Evolving from "AI that can chat" to "AI that can get work done" has become the concentrated manifestation of current AI iteration.

Sutton pointed out that the stock of manually labeled resources continues to decline and cannot support long-term model iteration. Agents, by contrast, can learn autonomously based on observations, actions, and reward signals generated through their interaction with the environment, with data expanding synchronously with their own capabilities. Both AlphaGo and the mathematical Olympiad reasoning system AlphaProof have confirmed the feasibility of this path. Existing large models lack goals, rewards, and punishment feedback, making them unable to independently verify the authenticity of information; experience-driven agents can continuously correct their cognition relying on environmental feedback, and autonomous evolution is an irreversible technological trend.

Tencent Research Institute, in collaboration with Tencent Cloud Intelligence and Tencent Youtu, released the "Co-evolution: 2026 Top 10 Trends in Artificial Intelligence Report" (hereinafter referred to as the "Report") at 2026 WAIC. The report argues that "Intelligence as a Service" is becoming a new direction for business model upgrading. In the past, enterprises obtained professional capabilities mainly by recruiting personnel, purchasing software, or outsourcing services; now they can directly call mature AI intelligence services, where underlying computing power consumption is no longer the focus of attention, and deliverable, verifiable practical results have become the core of value.

The report summarizes this new network form driven by AI agents as the "Intelligence of Everything". That is to say, AI agents are no longer just tools used by humans, but new participants on the internet, capable of independently calling services, completing tasks, and carrying out multi-agent collaboration. Correspondingly, the commercial logic of the internet has gradually shifted from the past attention-oriented model to a practical result-oriented one, and the core of platform competition has turned to capability scheduling and task delivery efficiency.

However, Zhan Bingqiang, Founder of AIGCLink, said in an interview with the *Time Weekly* reporter that the combination of agents, software and hardware is only the surface of competition. For major tech giants, staying in the game still fundamentally depends on model capabilities.

"Currently, competition among major giants is concentrated on the client side, which is essentially a move to seize user data. The accumulation of user data helps giants iterate their own models. If the model capability is not strong, no matter how well the client is built, it may still be impacted by model iteration. Just like OpenClaw, which was once a hit in the first half of this year."

Zhan Bingqiang believes that the landscape of major tech giants may change in the future. The leading players may not necessarily be Alibaba, ByteDance, or Tencent, but could also be large model companies such as Zhipu AI.

Zhu Yali, General Manager of Wisdom Bud China, told the *Time Weekly* reporter in an interview that the path competition among tech giants is essentially a matter of choosing directions: to pursue "the most general capability" or "the most certain result". The general-purpose route attempts to solve all problems with one model, covering a wide range of areas, but tends to stop at superficial results once entering professional scenarios; while the vertical route focuses on specific industries, using domain data and business know-how to thoroughly solve a certain type of problem.

Major tech giants have clearly realized this issue. For example, the WorkBuddy family is very large. It not only has a version for workplace people, but also includes WorkBuddy Enterprise Edition and the office agent suite Agent Suite for enterprise users, providing differentiated and personalized agent solutions for different groups of people and more than 20 vertical scenarios.

In Zhu Yali's view, many current agents are still limited to single-point efficiency improvement. "I am more optimistic about another path: making agents responsible for business results, and finally precipitating them into organizational capabilities. To measure the success of AI implementation, we should not only look at how much efficiency has been improved in a certain position, but also see whether the innovation capability of the entire organization has been elevated to a new level due to AI."

Tang Jie, Founder of Zhipu AI, issued an internal letter titled "The Great Wave Has Arrived" on July 11, pointing out that there is a clear path for the evolution of the upper bound of intelligence. Artificial intelligence is completing the leap from perceptual intelligence to cognitive intelligence. Machines are no longer just "seeing" and "hearing", but beginning to "understand" and "reason". And the next step is heading towards AGI.

(Our reporter Xie Silin also contributed to this article)

This article is from the WeChat official account "Time Weekly" (ID: timeweekly), author: Wang Ying, published with authorization from 36Kr.