AsiaInfo Technologies (亚信科技)

AI can automatically extract rules and knowledge from historical code, API assets, and architecture characteristics, then preserve daily development conversations as knowledge cards. (original: AI 能够从海量历史代码、API 资产和架构特征中自动提取规范和知识，并将日常研发对话沉淀为知识卡片。)

After three years of AI transformation, AsiaInfo brought nearly ten thousand R&D staff into an AI Coding era and turned senior experts' tacit experience into reusable team capability. (original: 通过为期三年的 AI 转型实践，亚信科技成功将近万名研发人员带入“全员 AI Coding”时代。Qoder 的引入，不仅将老专家的隐性经验转化为可复用的团队能力。)

AsiaInfo said its BSS and OSS systems span multiple generations from 2G to 5G, with intertwined old and new logic and high module coupling that makes R&D and maintenance extremely difficult. (original: 系统横跨从 2G 到 5G 的多个技术代际，新老业务逻辑交织，模块间耦合度高，“牵一发而动全身”，研发与维护难度极大。)

Average code output per engineer increased by about 14.6%, and average end-to-end efficiency from requirements through testing improved 8.6%. (original: 人均代码产出提升约14.6%；从需求到测试的端到端全流程平均提效8.6%。)

For complex defect work, AsiaInfo uses Qoder to run multiple agents in parallel for localization, repair, and verification while they share context and are recomposed at the architecture layer. (original: 将复杂的业务需求拆解给缺陷定位、修复、验证等多个 Agent 并发执行。Agent 之间共享上下文并互相协同，最后在架构层进行合成。)

Overall AI coverage of core BSS systems reached about 80%, and monthly active rate stayed above 90% in a 1,000-person operations team. (original: 核心 BSS 系统 AI 覆盖率整体达到80%左右；在千人规模的运营团队中，月活跃率稳定90%以上。)

Qoder was rolled out across the full R&D system, not just code completion, progressively covering requirements, design, testing, and defect repair. (original: 通过 Qoder，亚信科技不仅实现了代码补全，更逐步覆盖了需求、设计、测试、缺陷修复等完整的研发体系。)