Jiang Guanjun, head of AI algorithms at Alibaba's Qwen business unit, set out the framework on October 22 at the 2026 China National Computer Congress in Chengdu.
Jiang split AI application development into three stages: chat, where AI answers questions; agent, where it completes tasks; and assistant, where it combines professional knowledge with personal context and spots unstated needs.
He argued that a personal agent could theoretically solve close to 100% of everyday problems, with the bottleneck being matching system capabilities to the right path for each problem.
On memory, Jiang said a personal agent must understand a user over the long term without piling up chat history, which adds noise and cost, and Qwen designs memory around how strongly events relate to each other.