DingTalk has recently established a new business line focused on industry-specific models, operating as an independent team that reports directly to DingTalk CTO Zhu Hong. This represents a significant strategic move in DingTalk's AI advancement following the return of founder Wuzhao.

According to sources familiar with the matter, "Since the team's formation, DingTalk has engaged with multiple industry clients, and several industry/enterprise-specific models are currently in development."

Since returning to DingTalk in April, Wuzhao has prioritized product experience and AI innovation as top priorities. Starting in April, DingTalk underwent comprehensive reforms covering product design, reviews, and improvements, with Wuzhao deeply involved in frontline operations.

Following ChatGPT's explosive popularity, DingTalk has completed integration of foundational large model capabilities. In August 2023, DingTalk opened its intelligent infrastructure (AI PaaS) to ecosystem partners and customers, encouraging partners to leverage large models to rebuild their products. In January 2024, DingTalk launched an AI assistant with perception, memory, planning, and action capabilities that can execute tasks across applications.

According to previously disclosed figures, DingTalk currently serves over 25 million enterprise organizations, with more than 2.2 million enterprises using AI on the platform, covering 20 primary industries including manufacturing, healthcare, finance, and retail.

The establishment of the industry-specific model team represents continued enterprise-side implementation following large model advances in technology and productization.

Enterprise AI implementation faces significant challenges. On one hand, most enterprises, particularly small and medium-sized businesses, have strong AI demands but generally lack professional technical teams and data processing capabilities. On the other hand, while general large models are powerful, they struggle to meet specialized vertical industry needs, requiring deep customization and optimization for specific scenarios.

The DingTalk industry-specific model team primarily serves enterprise customers and third-party partners on the DingTalk platform. For small and medium enterprises lacking sufficient AI talent resources, DingTalk provides end-to-end model training and data engineering services, including frontend data labeling, cleaning, and model optimization, all completed by DingTalk teams.

Compared to developer-focused platforms like Alibaba Cloud's Bailian, DingTalk's industry-specific models are more closely aligned with business scenarios.

"Industry-specific models are co-created by DingTalk and business personnel within enterprises who understand business operations and industry specifics, allowing industry know-how to be preserved and enabling enterprise customers to adopt models more quickly and effectively," a DingTalk representative stated.

DingTalk has already achieved preliminary results in industry-specific models. The Doukou Gynecology Large Model released in July represents the first successful vertical-specific large model deployed on the DingTalk platform. As a healthcare vertical model, it improved diagnostic accuracy for six major gynecological symptoms from 77.1% to 90.2%.

While advancing industry-specific models, DingTalk is also accelerating completion of its AI ecosystem loop, with another new initiative being application marketplace revisions.

Agents have become an open competitive battleground for major tech companies in 2025. Starting in 2024, major players have launched Agent platforms including Alibaba Cloud Bailian, ByteDance's Coze, Baidu's Wenxin Intelligent Agent, and Tencent's Yuanqi. DingTalk also launched its AI Agent Store in April 2024.

Following Wuzhao's return to DingTalk, an important focus has been reconstructing the Agent marketplace logic. Beyond significant changes to application recommendation methods, DingTalk plans to further develop the Agent marketplace, opening capabilities to more ISVs and enterprises, helping companies build Agent applications, and achieving commercial loop closure through DingTalk to connect the entire Agent application ecosystem.