AI Trends

ByteDance's Coze Enterprise Edition to Be Phased Out

East MoneyUpdated Nov 6 10:34

AI summary

ByteDance's AI application development platform Coze has announced that it will phase out its Enterprise Edition: new sales of the Enterprise Standard and Flagship editions will stop starting November 30, 2026, legacy plans will be discontinued on December 31, and the new enterprise plans will be fully discontinued by the end of 2027 at the latest. In August this year, the Coze team was merged into the Doubao system; people familiar with the matter said that its target users and functions overlap with Doubao Work, making product integration an inevitable trend. Tencent Yuanqi will cease service on November 9, and Baidu's Miaoda team has been merged into Baidu Dazi. The entry points of AI application development platforms are being squeezed by AI office products such as Doubao Work and Qianwen Office.

Why read this

AI score 60

From the phasing out of Coze Enterprise Edition, the integration trend of AI application development platforms being replaced by office Agents, and the remaining space for independent development platforms in complex enterprise scenarios.

Topics

Tags

Body · AI translation

Article body

Coze, the AI application development platform under ByteDance, has issued an announcement stating that it will phase out its Enterprise Edition. The Enterprise Standard and Flagship editions will stop new sales starting November 30, 2026, the old plans will be taken offline on December 31, and the new enterprise plans will be fully discontinued by the end of 2027 at the latest.

Coze positions itself as a zero-code AI application development platform, focused on enabling ordinary people, developers, and business personnel to drag-and-drop orchestrate workflows, plugins, and multi-Agent systems to quickly build Bots, applications, and automated tasks, which can be published externally and integrated via API. Coze launched its overseas version in December 2023 and released its domestic version two months later.

In August of this year, the Coze team was already merged into the Doubao system. A person familiar with the matter told Jiemian News that ByteDance has internally made Doubao Work a core product in the AI track, and that the Coze Enterprise Edition's target users and functions overlap to some extent with Doubao Work, making product integration an inevitable trend.

Prior to this, ByteDance had also carried out multiple rounds of integration of AI productivity products. Among them, the Feishu product team was merged into the Doubao system, Feishu's original built-in AI assistant Aily was renamed Doubao Work Companion, and ByteDance's AI programming product TRAE was also merged into the Doubao system.

Not only ByteDance, but internet giants such as Alibaba, Tencent, and Baidu also integrated their AI productivity products this year. Among them, Tencent's AI application development platform Yuanqi will cease service on November 9, and the operations team of Baidu's zero-code AI application development platform Miaoda has been merged into Baidu Dazi. A Baidu insider revealed to Jiemian News that after this adjustment, Baidu Dazi will serve as the core entry point for its AI office products, while Miaoda will exist as a Skill within Baidu Dazi.

On the one hand, the traffic entry points for AI productivity products are concentrating toward low-threshold tools. In the early days of large models, users of AI application development platforms were mainly frontline R&D personnel, and the products had a high usage threshold: users not only had to find them, but also had to learn to configure workflows and knowledge bases before they could start using them. Today, after the emergence of AI products such as Workbuddy, Doubao Work, and Qianwen Office, Agents have stronger capabilities and the user threshold has been greatly lowered, completing the replacement of AI application development platforms.

On the other hand, AI application development platforms were born in a period when model capabilities were not yet strong enough, and their previous product designs also contained a large number of modular and process-based configuration steps. For example, Coze breaks AI applications down into different functional components such as models, knowledge bases, plugins, and memory, and provides a visual workflow canvas. Users can drag different nodes to break tasks down into multiple steps and define the input, output, and execution conditions of each step. Today, long context, autonomous planning, tool calling, and multi-step execution have become basic capabilities of models, and users only need one sentence to let an Agent break down tasks, find tools, call systems, and write back results on its own.

After trying it out, Jiemian News found that Coze's current drag-and-drop modular canvas is no longer the main promoted route; the product's focus has fully shifted to Coze Programming, prioritizing a natural language generation plus code-first paradigm, and it is no longer very different from Doubao Work, with the overlap between the two already very high.

The above-mentioned industry practitioner said that the feature Coze was considered to do very well at the time was that Bots (knowledge base + workflow + plugins) required no additional development and no self-written integration code, and could be deployed directly to platforms such as Feishu, Douyin, and Doubao after authorization. But today these entry points have already grown the ability to generate applications.

In the future, most ordinary enterprise employees will no longer need to care about how an Agent is built; they will only need to care about whether it can solve problems. When employees raise requirements in an AI office platform, the Agent can automatically call knowledge bases, plugins, and enterprise systems according to the task, completing the entire process from task breakdown to result delivery.

This will lead to further concentration of AI application entry points. Relying on the users, organizational relationships, and business scenarios accumulated by office platforms such as WeCom, Feishu, and DingTalk, Workbuddy, Doubao Work, and Qianwen Office will have the opportunity to become the unified entry point for enterprises to call AI capabilities. And the application building, publishing, and usage steps that were previously scattered across independent development platforms will gradually be integrated into these AI office platforms.

But this does not mean that independent AI application development platforms will completely lose their room to survive. The above-mentioned industry practitioner said that for large enterprises with complex business processes, private deployment, and deep system integration needs, open development platforms still have value. However, the focus of competition is no longer how many draggable components are provided, but who can better connect models, data, and enterprise systems and solve the problem of Agents running stably in real business.

And the competition in enterprise-level Agents has another deeper variable: enterprise context. Chen Yusen, CEO of Qianwen Office and Vice President of Alibaba Group, gave a clear judgment at this year's Yunqi Conference: Context is All You Need. In his view, the key to the real implementation of enterprise-level Agents is not just model capability, but data. Even if a model is smart enough, it cannot know out of thin air the unique business rules and ways of working of an enterprise.

But building enterprise context is not easy. Gao Ting, Vice President of Research at Gartner, told Jiemian News that there is currently no fully mature one-stop enterprise context solution on the market, and enterprises need to use some modular technical solutions to build their own AI context platforms.

This is also the opportunity for independent AI application development platforms. AI office platforms can enable Agents to quickly enter employees' daily work. The potential value of independent development platforms lies in more open and flexible development capabilities, allowing enterprises to independently combine models, data, tools, and workflows according to their own technical architecture and business needs.

(Source: Jiemian News) Middle of the body Embedded advertisement

Source

East Money