ByteDance announced a major organizational restructuring on July 30, merging its Doubao AI assistant with Feishu productivity software to create a unified enterprise AI platform. The move signals a strategic shift from internal competition to coordinated product development, as Chinese tech giants race to dominate the lucrative business-to-business AI market.

Why Is ByteDance Consolidating Its AI Products Now?

The restructuring reflects a broader industry trend. Over the past month, Alibaba integrated three Agent products into "Qianwen Work," while Tencent merged its QClaw business into the WorkBuddy system. Huawei took an even more integrated approach, making Xiaoyi the unified AI entrance for all HarmonyOS scenarios from the start. ByteDance cannot afford to fall behind, as the competitive window for desktop AI office tools is closing rapidly.

According to Q2 2026 market data, Tencent's WorkBuddy leads with 20.97 million monthly visits on desktop, while ByteDance's domestic Trae ranks second with 12.79 million visits. The total monthly visits across 17 mainstream domestic AI office agents exceeded 60 million in June. Desktop-level productivity demands, such as writing long documents and processing complex spreadsheets, represent far more valuable engagement than mobile clicks.

How Will the Merged Structure Work?

• Product Integration: Zhao Qi, current head of Doubao, will lead the combined Doubao product team, with Feishu head Xie Xin reporting to him. Feishu's existing products and services remain unchanged, but will deepen collaboration with Doubao in productivity scenarios.
• Sales and Marketing Consolidation: Feishu's go-to-market team merges with Volcano Engine's team to form a new "Creativity Service Platform" led by Tan Dai. This unified organization handles marketing, sales, and customer service for ByteDance's cloud services, including Model-as-a-Service (MaaS) and Software-as-a-Service (SaaS) offerings.
• Enterprise Edition Launch: Doubao Enterprise Edition, developed with Feishu's participation, has begun internal beta testing. It natively integrates Feishu's ecosystem, allowing users to connect office scenarios like Feishu Docs, Sheets, Meetings, and group chats without additional deployment, plus enterprise-grade data isolation and security auditing.

This represents the largest adjustment to ByteDance's business-to-business operations since the company implemented business segmentation in 2021. The consolidation moves ByteDance from what executives call an "internal horse racing period" of scattered AI investments to a "entrance competition period" where resources concentrate on leading products.

What Are the Financial Pressures Behind This Move?

ByteDance faces mounting costs from its AI ambitions. The company's net profit in 2025 fell more than 70 percent year-over-year, primarily due to substantial increases in capital expenditure on AI computing power, model research and development, and data center infrastructure in the third and fourth quarters of 2025. ByteDance has raised its AI infrastructure capital expenditure for 2026 to more than 200 billion yuan (approximately $28 billion USD).

Despite Doubao's dominant consumer position, with daily active users exceeding 200 million after the 2026 Spring Festival and monthly active users reaching 382 million, the model's profitability remains elusive. Even ChatGPT, with over 900 million weekly active users and more than 50 million paying subscribers, achieves only a 5 percent paid conversion rate.

Feishu, by contrast, offers a proven monetization path. The collaboration software generated annual recurring revenue exceeding 3 billion yuan in 2025, with second-quarter 2026 revenue increasing more than 100 percent year-over-year. Notably, 90 percent of Feishu's new customers in 2026 purchased Feishu AI products simultaneously, positioning the platform to rank first in domestic enterprise collaboration revenue within 2026.

"Coding, as a highly structured and logically rigorous task, has high requirements for the model to understand complex semantic structures, logical reasoning, algorithm design and accurate expression, and can help explore the upper limit of model intelligence," said Hong Dingkun, Vice President of ByteDance Technology.

Hong Dingkun, Vice President of ByteDance Technology

How Does This Fit Into the Broader AI Coding Competition?

Beyond enterprise productivity, ByteDance is also competing fiercely in AI coding, another high-value market segment. Anthropic's Claude Code generated annualized revenue exceeding 1 billion US dollars in just six months after launch in February 2025, demonstrating the commercial potential of AI-assisted programming. OpenAI's Codex has gained momentum, with usage approaching 40 percent of Claude Code's levels by January 2026, up from just 5 percent in September 2025.

Alibaba's Qoder, launched less than a year ago, captured 47.6 percent of the AI programming revenue share according to IDC data, signaling that domestic companies are rapidly closing the gap with Western competitors. ByteDance's Trae, Tencent's CodeBuddy, and other domestic platforms are all accelerating their coding capabilities as companies recognize that AI coding is no longer just a productivity tool but a foundation for determining the upper limit of AI agent effectiveness.

The Doubao large model itself is developed by ByteDance's Seed team, which currently employs more than 2,000 people, including core researchers, infrastructure engineers, data labeling teams, and translators. Led by Dr. Wu Yonghui, the Seed team focuses on fundamental large language model research and reports directly to CEO Liang Rubo. Maintaining Doubao's leading position in the consumer market remains the starting point for all ByteDance business strategies.

By embedding Doubao's AI capabilities into Feishu and distributing them through Volcano Engine's cloud service channels, ByteDance aims to convert its proven business-to-business monetization capabilities into a vehicle for its consumer AI technology. This strategic consolidation reflects the industry's recognition that the "internal horse racing period" has ended, and the race for enterprise AI dominance is now in full acceleration.