AI Version Alipay "Treasure Plan" Details Unveiled: Pioneering Interface + Screen Reading Dual-Track System to Mobilize Mini Programs

According to Beating monitoring, Alipay’s internal codename for its AI redesign is “Treasure Plan,” and the project is currently personally led by Li Jun, President of the Alipay Business Group. According to details disclosed by LatePost, as Ant Group’s highest-classified project, the new version is expected to go live soon. After release, it will undergo rapid iterations at a frequency of twice per month. To serve the existing 1 billion user base and achieve zero migration cost, the new version adopts a “treadle-and-bird” switching model—once users turn it on, it defaults to the original Alipay, while also letting users independently switch into a brand-new conversation-driven interface with a single click and set it as their preferred option.

On the technical side, the new version uses a dual-track approach. On one hand, it drives merchants to transform their services into MCP interfaces that AI can directly call. On the other hand, with user authorization, it uses screen-reading operations to be compatible with mini programs that have not yet been transformed, thereby completing a closed-loop service for long tasks. Alipay also plans to launch an AI open platform after the rollout and promote cross-end interconnection across different terminals such as mobile phones, in-car systems, and smart glasses.

In terms of strategic evolution, the redesign project has gone through major route adjustments. When the project team was established in the second half of 2023, it chose to develop an independent native AI application at the outset, partly to avoid the complexity of intertwining with the original terminal business. Then, in September 2024, it released Xiao Bao. However, the standalone app’s results did not meet expectations, while the daily active user count of the intelligent agents within Alipay has remained steady at the level of several million. Due to the excessively high costs of user acquisition and migration for the standalone app, and limited development resources for compute, in March 2025 the team made a directional course correction—abandoning the standalone app plan—and in December of the same year it formally approved the in-terminal transformation, gradually implementing the “treadle-and-bird” switching scheme.