AI-driven coding is being adopted across Ping An Group's core business lines, including banking, insurance, and technology. In some new projects, the proportion of AI-generated code has exceeded 70%.
Currently, more than 15,000 R&D engineers at Ping An Group are performing AI-assisted coding through the group's self-developed tool "Ping An Ai Ma", which integrates Alibaba's Qoder CN to enhance its code completion capabilities.
Launched in 2024, "Ping An Ai Ma" now offers core capabilities including intelligent code completion, intelligent Q&A, RAG-based knowledge management, and Agent+MCP task planning and execution.
It covers not only conventional R&D scenarios such as code generation and test data preparation—enabling AI-driven coding and coding-standard-compliant development—but has also achieved technical breakthroughs in specialized scenarios such as business platform construction and XML-to-SQL script conversion.
At present, the monthly volume of AI-generated code committed via "Ping An Ai Ma" continues to grow, with deep integration across the R&D lifecycle. It has delivered significant results across four core phases: requirements, development, testing, and review.
// Requirements
Intelligent story-splitting assistance and requirement-specification self-checks standardize requirement granularity. All solution designs can be produced rapidly based on requirements, with efficiency improved by more than 70%.
// Development
Through the deep integration of prompt engineering, RAG, Agent technology, and MCP, high-value scenario coverage has been expanded, progressively achieving automation across R&D scenarios. For example, in unit testing, AI assists in completing 90% of the writing work, improving efficiency by more than 80%.
// Testing
Centered on four flagship capabilities—functional test case generation, AI-based API automation test case generation, AI-driven intelligent data fabrication, and AI-powered precision analysis—hundreds of thousands of test cases have been committed to the repository, substantially improving testing efficiency.
// Review
AI-driven requirement review and code review gating enforce requirement standardization and raise the baseline for code quality, safeguarding R&D efficiency.
The head of the "Ping An Ai Ma" team said: "We have infused project experience, management standards, and expert knowledge into 'Ping An Ai Ma', and we will continue to iterate on the latest large models including MCP, agents, and Qwen3-Coder. Our goal is not merely to use 'Ping An Ai Ma' as an assistive tool, but to bring about transformation and innovation in R&D paradigms—accelerating Ping An Group's AI-driven transformation at full speed."
Customer Cases