A fierce competitive battle has erupted in China's office artificial intelligence market, where Baidu's Dazi agent has taken leading positions in growth rates. According to the July AICPB rating, Dazi reached 6.74 million monthly active users on desktop computers. It only trailed Tencent WorkBuddy, which has 11.15 million users, but demonstrated a growth of 1063.79% month-over-month, the highest figure among all office AI agents.
During a product presentation on August 27, Baidu announced that the number of Dazi users increased nearly ninefold in the last month. Since its launch in March, the agent has undergone about 150 iterations, sometimes releasing updates almost daily. The current enterprise version includes 15 packages and 96 skills covering areas such as finance, law, product development and R&D, operations, and human resources. The number of daily queries since launch has grown 60-fold, and website traffic in July increased by 845% compared to the previous month.
The company attributes Dazi's advantage to organizational factors rather than purely algorithmic ones. Shen Dou, Vice President of Executive Affairs, asserts that an effective agent must not only possess a superior model but also deeply understand the user and be able to manage tools and professional methods. Dazi leverages the benefits of Baidu's full-featured cloud AI and its ability to customize delivery, which its competitors—Tencent's WorkBuddy, focused on communication and documents, and Alibaba's Qwen, focused on teamwork—reproduce less easily.
Behind the product is a broader strategic vision from Chairman Robin Li. At the AIDAY conference in June, he proposed the 'DAA' (Daily Active Agents) metric as a new measure of AI value, accounting for how many agents actually perform tasks daily, rather than just the volume of tokens consumed. In August, he outlined a course for ERNIE based on 'application,' suggesting strengthening through real, most important applications, including AI search, digital humans, Miaoda, Famou, and Dazi.
This strategy represents a self-sustaining cycle: real products define the models of what capabilities are significant; the model optimizes the execution of these tasks; adoption levels grow; and new usage data feeds into the next cycle. As functions converge in the industry, Baidu is betting that whoever implements this 'application-oriented' mechanism more deeply will become the leader in office AI that users will genuinely use.

