The first AICPB AI Product Ranking desktop list presents China’s AI office market as both sizable and unsettled. In July 2026, AI office products running on domestic PC clients recorded about 30 million monthly active users (MAU). Tencent’s WorkBuddy led by audience size, while Baidu Dazi was the category’s fastest-growing product. 1

The 15-second takeaway

The biggest signal is not simply that one product is ahead. It is that AI office software is emerging as a distinct platform battleground.

The leading products are expanding by connecting AI to documents, email, cloud drives, meetings, search and cross-application task execution. Dozens of products are already competing for users, however, so the current ranking should not be treated as a settled market structure. 1

WorkBuddy wins on scale; Dazi wins on speed

Tencent WorkBuddy ranked first with 11.1523 million MAU, up 304.40% month on month. Baidu Dazi ranked second with 6.7430 million MAU, but its 1,063.79% month-on-month growth put it at the top of the growth table. 1

Together, the two products had nearly 18 million MAU, or about 60% of the estimated desktop AI office market. That points to clear concentration at the top, while leaving roughly 40% of users distributed among other products. 1

Product	July desktop MAU	Month-on-month change
Tencent WorkBuddy	11.1523 million	+304.40%
Baidu Dazi	6.7430 million	+1,063.79%
TraeWork	2.6599 million	+51.08%
Coze	2.1733 million	+62.23%

Source: AICPB’s first July 2026 desktop ranking for AI office agents. 1

ByteDance is spreading its office bet across several products

ByteDance’s TraeWork ranked third with 2.6599 million MAU, up 51.08%, while Coze ranked fourth with 2.1733 million MAU, up 62.23%. 1

According to AICPB-related reporting, ByteDance is building a broader office-product portfolio rather than relying on a single assistant. The Feishu product team was reportedly brought into Doubao, with work-task features added and capabilities such as Feishu Sheets connected. TraeWork and Coze have also been emphasized, partly because Doubao’s original consumer and entertainment positioning may not naturally support complex office tasks. 1

The strategy reflects a wider shift in AI office competition. Products do not necessarily need to evolve from a general-purpose chatbot into workplace software. They can also attract users through more specific entry points, including coding, task execution, collaboration and tool calling.

The common playbook: connect AI to existing workflows

Baidu Dazi: using search and content capabilities as a starting point

Baidu launched Dazi in March 2026 and integrated capabilities from Baidu Baike, Baidu Maps, Miaoda and Baidu Netdisk into the office-agent platform. By July, Dazi had reached 6.7430 million MAU and recorded 1,063.79% month-on-month growth. 1

Its rise suggests that the value of an office agent lies in more than text generation. A useful product must be able to interpret a work objective, break it into steps, retrieve information, process files and call multiple applications in sequence. 1

Tencent WorkBuddy: expanding through an established office ecosystem

WorkBuddy began as a small-team initiative before receiving additional resources and connecting with Tencent Docs, QQ Mail, Tencent Netdisk and Tencent Meeting. Its positioning is closer to an AI workbench spanning documents, email, collaboration and meetings. 1

This kind of integration may reduce switching costs. Users can add AI capabilities to services they already use instead of rebuilding their entire work routine around a new platform.

Alibaba: consolidating QwenWork, Wukong and other projects

At the end of July, Alibaba integrated QwenWork with Wukong, which was incubated by DingTalk, along with the internal venture MuleRun and QcodeWork. The ranking treated the move as part of Alibaba’s broader effort to organize its AI office products and capabilities. 1

While the market has yet to settle on a standard product format, combining model capabilities, workplace collaboration channels and task-execution tools may offer a more direct path into real workflows than launching another standalone chat interface.

A second tier is already taking shape

Beyond the top four, AICPB listed the following July desktop MAU figures for competing AI office products: 1

• QcodeWork: 1.8855 million
• QwenWork: 607,900
• Wukong: 588,500
• LobsterAI: 511,100
• Xiaohuanxiong: 379,700
• easyclaw: 274,900
• Lingxi Pro: 258,200
• Nano Work: 227,300
• AutoClaw: 143,500
• Loomy: 7,400
• krowork: 3,000
• AstronClaw: 1,400

The figures show that the market is not limited to a three-way contest among Tencent, Baidu and ByteDance. Products may be building audiences around coding, automation, enterprise collaboration or specialized tasks. The ranking itself does not provide enough information to assess retention quality or commercial performance. 1

What does 30 million desktop MAU mean?

AI office has moved beyond the proof-of-concept stage

About 30 million monthly active PC users indicates that AI office tools already have a substantial base of real users, rather than existing only as enterprise pilots or novelty products. AICPB also suggested that the category could have at least 20 times the growth potential of traditional office software. 1

That “20 times” figure is the ranking publisher’s estimate of potential, not a guaranteed market forecast. Future growth will depend on whether products can reliably handle complex tasks, whether companies adopt them and whether users return after an initial trial.

Concentration and variety are developing at the same time

WorkBuddy and Dazi’s combined share shows the importance of distribution, existing ecosystems and investment. At the same time, TraeWork, Coze, QcodeWork and a range of newer products have attracted meaningful desktop audiences, suggesting that users are not yet locked into one platform. 1

The next phase of competition may therefore depend on a combination of capabilities:

1. Access to real work materials: Can the product handle documents, email, spreadsheets and cloud files?
2. Cross-application execution: Can it move beyond understanding a request to researching, organizing and taking action?
3. Integration with collaboration systems: Can it participate in meetings, projects and team workflows?
4. Manageable costs: Can it balance quality, speed and token consumption?

More competition could lower user costs

AICPB argued that intensifying competition may lead to faster product iteration, lower token costs or more generous token allowances. 1 For users, that makes actual task performance more important than a model’s brand name or marketing feature list.

DAU at 20% of MAU? Treat it as a rough benchmark

The ranking did not publish daily active user (DAU) figures for each product. Instead, it suggested that a well-performing AI office agent might have DAU equal to roughly 20% of MAU. 1

This is an informal estimate, not a measured result for every product and not a substitute for retention data. MAU records whether someone used a product during a month, while DAU reflects daily usage frequency. Assessing product quality also requires metrics such as task-completion rates, repeat usage, enterprise renewals and user retention.

Important limitations of the ranking

AICPB describes its list as a reference ranking based on publicly available online data and highlights several limitations: 1

• Product classification may not be fully accurate.
• Coverage may be incomplete.
• The ranking is not an endorsement or a compliance assessment.
• It should not be used as investment, employment or other decision-making advice.
• AICPB does not guarantee the data’s accuracy, reliability or completeness.

The figures cover domestic PC-client MAU. They cannot be directly compared with mobile MAU, website traffic or global AI rankings. Differences in measurement methods, installation channels and user-identification practices may also affect the results. 1

For non-commercial republication or use of the data, AICPB asks users to credit the source as “AI产品榜 AICPB.” 1

Bottom line: the workplace is becoming AI’s next major battleground

The first desktop ranking points to three conclusions: the market already has scale, growth remains rapid, and the eventual winners will be the products that embed AI reliably into real workflows.

China’s overall AI user base was estimated at about 600 million monthly active users, making the workplace an entry point that major platforms are unlikely to ignore. 1 But the 30 million desktop MAU figure and the leaders’ rapid growth are better understood as signs that competition is entering a larger phase—not proof that the market’s final hierarchy has been decided.

For businesses and individual users, the most important question is no longer simply which product tops the chart. It is whether these office agents can consistently complete useful work, across the tools and files people already depend on.

The first AICPB AI Product Ranking desktop list presents China’s AI office market as both sizable and unsettled. In July 2026, AI office products running on domestic PC clients recorded about 30 million monthly active users (MAU). Tencent’s WorkBuddy led by audience size, while Baidu Dazi was the category’s fastest-growing product. 1

The 15-second takeaway

The biggest signal is not simply that one product is ahead. It is that AI office software is emerging as a distinct platform battleground.

The leading products are expanding by connecting AI to documents, email, cloud drives, meetings, search and cross-application task execution. Dozens of products are already competing for users, however, so the current ranking should not be treated as a settled market structure. 1

WorkBuddy wins on scale; Dazi wins on speed

Tencent WorkBuddy ranked first with 11.1523 million MAU, up 304.40% month on month. Baidu Dazi ranked second with 6.7430 million MAU, but its 1,063.79% month-on-month growth put it at the top of the growth table. 1

Together, the two products had nearly 18 million MAU, or about 60% of the estimated desktop AI office market. That points to clear concentration at the top, while leaving roughly 40% of users distributed among other products. 1

Product	July desktop MAU	Month-on-month change
Tencent WorkBuddy	11.1523 million	+304.40%
Baidu Dazi	6.7430 million	+1,063.79%
TraeWork	2.6599 million	+51.08%
Coze	2.1733 million	+62.23%

Source: AICPB’s first July 2026 desktop ranking for AI office agents. 1

ByteDance is spreading its office bet across several products

ByteDance’s TraeWork ranked third with 2.6599 million MAU, up 51.08%, while Coze ranked fourth with 2.1733 million MAU, up 62.23%. 1

According to AICPB-related reporting, ByteDance is building a broader office-product portfolio rather than relying on a single assistant. The Feishu product team was reportedly brought into Doubao, with work-task features added and capabilities such as Feishu Sheets connected. TraeWork and Coze have also been emphasized, partly because Doubao’s original consumer and entertainment positioning may not naturally support complex office tasks. 1

The strategy reflects a wider shift in AI office competition. Products do not necessarily need to evolve from a general-purpose chatbot into workplace software. They can also attract users through more specific entry points, including coding, task execution, collaboration and tool calling.

The common playbook: connect AI to existing workflows

Baidu Dazi: using search and content capabilities as a starting point

Baidu launched Dazi in March 2026 and integrated capabilities from Baidu Baike, Baidu Maps, Miaoda and Baidu Netdisk into the office-agent platform. By July, Dazi had reached 6.7430 million MAU and recorded 1,063.79% month-on-month growth. 1

Its rise suggests that the value of an office agent lies in more than text generation. A useful product must be able to interpret a work objective, break it into steps, retrieve information, process files and call multiple applications in sequence. 1

Tencent WorkBuddy: expanding through an established office ecosystem

WorkBuddy began as a small-team initiative before receiving additional resources and connecting with Tencent Docs, QQ Mail, Tencent Netdisk and Tencent Meeting. Its positioning is closer to an AI workbench spanning documents, email, collaboration and meetings. 1

This kind of integration may reduce switching costs. Users can add AI capabilities to services they already use instead of rebuilding their entire work routine around a new platform.

Alibaba: consolidating QwenWork, Wukong and other projects

At the end of July, Alibaba integrated QwenWork with Wukong, which was incubated by DingTalk, along with the internal venture MuleRun and QcodeWork. The ranking treated the move as part of Alibaba’s broader effort to organize its AI office products and capabilities. 1

While the market has yet to settle on a standard product format, combining model capabilities, workplace collaboration channels and task-execution tools may offer a more direct path into real workflows than launching another standalone chat interface.

A second tier is already taking shape

Beyond the top four, AICPB listed the following July desktop MAU figures for competing AI office products: 1

• QcodeWork: 1.8855 million
• QwenWork: 607,900
• Wukong: 588,500
• LobsterAI: 511,100
• Xiaohuanxiong: 379,700
• easyclaw: 274,900
• Lingxi Pro: 258,200
• Nano Work: 227,300
• AutoClaw: 143,500
• Loomy: 7,400
• krowork: 3,000
• AstronClaw: 1,400

The figures show that the market is not limited to a three-way contest among Tencent, Baidu and ByteDance. Products may be building audiences around coding, automation, enterprise collaboration or specialized tasks. The ranking itself does not provide enough information to assess retention quality or commercial performance. 1

What does 30 million desktop MAU mean?

AI office has moved beyond the proof-of-concept stage

About 30 million monthly active PC users indicates that AI office tools already have a substantial base of real users, rather than existing only as enterprise pilots or novelty products. AICPB also suggested that the category could have at least 20 times the growth potential of traditional office software. 1

That “20 times” figure is the ranking publisher’s estimate of potential, not a guaranteed market forecast. Future growth will depend on whether products can reliably handle complex tasks, whether companies adopt them and whether users return after an initial trial.

Concentration and variety are developing at the same time

WorkBuddy and Dazi’s combined share shows the importance of distribution, existing ecosystems and investment. At the same time, TraeWork, Coze, QcodeWork and a range of newer products have attracted meaningful desktop audiences, suggesting that users are not yet locked into one platform. 1

The next phase of competition may therefore depend on a combination of capabilities:

1. Access to real work materials: Can the product handle documents, email, spreadsheets and cloud files?
2. Cross-application execution: Can it move beyond understanding a request to researching, organizing and taking action?
3. Integration with collaboration systems: Can it participate in meetings, projects and team workflows?
4. Manageable costs: Can it balance quality, speed and token consumption?

More competition could lower user costs

AICPB argued that intensifying competition may lead to faster product iteration, lower token costs or more generous token allowances. 1 For users, that makes actual task performance more important than a model’s brand name or marketing feature list.

DAU at 20% of MAU? Treat it as a rough benchmark

The ranking did not publish daily active user (DAU) figures for each product. Instead, it suggested that a well-performing AI office agent might have DAU equal to roughly 20% of MAU. 1

This is an informal estimate, not a measured result for every product and not a substitute for retention data. MAU records whether someone used a product during a month, while DAU reflects daily usage frequency. Assessing product quality also requires metrics such as task-completion rates, repeat usage, enterprise renewals and user retention.

Important limitations of the ranking

AICPB describes its list as a reference ranking based on publicly available online data and highlights several limitations: 1

• Product classification may not be fully accurate.
• Coverage may be incomplete.
• The ranking is not an endorsement or a compliance assessment.
• It should not be used as investment, employment or other decision-making advice.
• AICPB does not guarantee the data’s accuracy, reliability or completeness.

The figures cover domestic PC-client MAU. They cannot be directly compared with mobile MAU, website traffic or global AI rankings. Differences in measurement methods, installation channels and user-identification practices may also affect the results. 1

For non-commercial republication or use of the data, AICPB asks users to credit the source as “AI产品榜 AICPB.” 1

Bottom line: the workplace is becoming AI’s next major battleground

The first desktop ranking points to three conclusions: the market already has scale, growth remains rapid, and the eventual winners will be the products that embed AI reliably into real workflows.

China’s overall AI user base was estimated at about 600 million monthly active users, making the workplace an entry point that major platforms are unlikely to ignore. 1 But the 30 million desktop MAU figure and the leaders’ rapid growth are better understood as signs that competition is entering a larger phase—not proof that the market’s final hierarchy has been decided.

For businesses and individual users, the most important question is no longer simply which product tops the chart. It is whether these office agents can consistently complete useful work, across the tools and files people already depend on.