(Yicai) Sept. 7 -- Competition in the Chinese market for artificial intelligence office agents is moving beyond desktop productivity tools into enterprise workflows, where the spending potential and barriers to entry are higher. Providers that can get AI to understand and join in real business processes stand the best chance of winning the race.

Alibaba Group Holding, Tencent Holdings, Baidu, ByteDance, and other Chinese big-hitters in the technology domain have been releasing new AI office agent features and products since June and, in some cases, reshaping their organizations around them.

The battlefield is widening beyond products to the customer-service capabilities, ecosystems, and computing infrastructure of businesses. For example, Alibaba's QwenWork and Tencent's WorkBuddy have been integrated with their respective workplace messaging apps DingTalk and WeCom, while ByteDance has brought the Lark, Trae, and Coze teams under Doubao and launched Doubao Work.

Raising the productivity of individual employees is only the first step, according to Shu Junliang, vice president and head of product research at QwenWork. A company's overall workflow can still be constrained by interactions between staff and between them and AI agents, so agents must evolve into systems that can transform enterprise workflows and organizational structures, he told Yicai.

Enterprise Context

Rather than competing solely for users and traffic at the application layer, major tech players are focused on securing the resource agents need most when they enter an organization: enterprise context. Apart from individual work habits and documents, agents also need to understand aspects of an enterprise, such as organizational relationships, permissions, customer information, production, and financial data.

Enterprise context is a key area for the next development phase, Shu pointed out. Turning the business data, processes, organizational structures, and accumulated staff knowledge that companies have into information AI can understand is becoming increasingly critical, he said.

Agents must be able to identify an employee's role, understand a project's background and retrieve relevant materials, while AI capabilities need to be broken down into reusable components that can be embedded into existing enterprise systems, Shu said.

That strategy is already visible across the major platforms. Alibaba has integrated QwenWork with DingTalk, while Tencent has connected WorkBuddy with WeCom, Tencent Meeting, and other workplace services. ByteDance has strengthened the link between its foundation models and agents, while Alibaba plans to connect Qwen Office more closely with AI models and computing resources from Alibaba Cloud. These ecosystems could become decisive competitive assets.

The rivalry will not be limited to isolated comparisons of individual applications, but will evolve into a full-stack AI contest around models, computing power, cloud services, agents, collaborative software, and enterprise data.

Helping Hand

"Office agents are merely a phase in product evolution, and the market may experience new changes every three to six months," Shu said. In the latest phase, the focus is no longer on a single product, but on the ability to continuously assist clients of various sizes in achieving AI transformation, Shu noted.

For example, a key step in adopting AI at state-owned carmaker Changan Automobile is having its senior executives actually use the technology, with key leaders in subordinate departments driving implementation. 

“At Changan Auto, agents such as QwenWork are not just productivity tools for individual employees but also powerful instruments for building an AI-native organization," Liang Guangzhong, assistant president and general manager of the transformation and efficiency department, said to Yicai.

Companies are looking forward to the creative ways of working and productivity boost AI will bring. Agents can improve efficiency in certain work scenarios by several or even 10s of times, noted Jing Xiaofei, deputy GM of transformation and efficiency at Changan Auto. The company will conduct a more systematic analysis of costs and returns, with the goal of ensuring that the gains from higher efficiency and business benefits exceed investment, he added.

Combined monthly traffic to 17 desktop AI office agents in China exceeded 60 million in June, up from 20 million in March, according to statistics from Analysys.

Editors: Tang Shihua, Martin Kadiev