ByteDance’s AI chatbot Doubao is reportedly undergoing significant team adjustments, with its core general session team set to halve in size. This move comes after Doubao achieved a substantial milestone of over 200 million daily active users (DAU), positioning it among China’s top AI applications.
Streamlining for Efficiency and Monetization
According to reports from LatePost, the Doubao general session team, previously comprising around 50 individuals and responsible for the chatbot’s core conversational capabilities, including strategy and evaluation, is now reducing its headcount. This adjustment, driven by Doubao’s head Zhao Qi, is seen as a move to increase personnel efficiency. Zhao Qi, who has a long tenure at ByteDance and previously oversaw platforms like Guanzhang Middle Platform and the advertising platform Pangle, also played a key role in building ByteDance’s AI talent ecosystem. Since taking over the combined Feishu and Doubao product teams in July, he has focused on optimizing operational efficiency.
As part of this restructuring, some team members will be reassigned to other ByteDance ventures, including the commercialization efforts of Doubao itself and the Feishu collaboration platform. The remaining personnel will be laid off. Previously, Zhao Qi had also integrated the separate Doubao evaluation team into various business groups.
Impact on Model Development
The adjustments also extend to the Product Posttrain-Chat team, which is responsible for fine-tuning models from ByteDance’s large model department, Seed. This team, led by Zhu Wenjia, develops smaller models tailored for Doubao’s use based on Seed’s foundational models. With decreasing business needs in this specific area, some employees have been transferred to other departments, such as Horizon RL and Product Posttrain-Work. Reports also indicate a high turnover in leadership for the dialogue model integration, with the role changing hands twice within a single month.
Future Directions and User Base Leverage
Despite these personnel changes, ByteDance emphasizes that Doubao is not abandoning its focus on conversational AI. The team continues to work on enhancing dialogue quality. Around June, the post-training team consolidated several small models supporting Doubao’s chat functions into a single, more streamlined model.
Industry insiders suggest that with a massive user base of 200 million DAU already secured, ByteDance is in a strong position to explore new directions and rapidly iterate on product and algorithm advancements. This existing user base and accumulated technological capabilities are seen as critical assets for future growth and the validation of new strategies. While the success of Doubao’s initial phase was largely attributed to its conversational product strength, the company is now navigating the growing challenges and commercialization bottlenecks inherent in dialogue-focused AI products in the current market.
