NIO, AsiaInfo Adopt Volcano Engine TRAE for Enterprise AI Coding

NIO integrated TRAE into its smart cockpit R&D. The developer daily activity rate exceeded 70%, AI code adoption reached 90%, and AI code contribution accounted for 19.8% of committed code. Li Qingqiu, head of R&D efficiency at NIO's digital cockpit and software development department, said TRAE enables experience accumulation and collaboration, pushing R&D capability from individuals to the organization.

NIO applied AI not only to code generation but also to code review, issue localization, and knowledge management. The company built a three-tier AI code review system: low-frequency full-scale local review scans the entire repository for architectural and historical issues; high-frequency incremental local review checks new code before each commit; and online incremental review acts as a team gate to report genuine problems. For knowledge inheritance, AI connects code, documents, and history to generate project wikis, and continuously summarizes new issues and lessons into a knowledge base accessible via natural language. AI also helps locate complex bugs by combining broader code context and historical solutions; several long-standing issues were resolved this way.

One developer at NIO's vehicle connectivity team uses TRAE as his primary IDE, with token usage expanded to 150 million. Another developer in the smart systems team used TRAE to generate architecture docs, module code, and test code, completing a task in two weeks that was expected to take a month. NIO envisions a multi-layer AI development ecosystem—knowledge, skill, workflow, and agent layers—where agents become new basic units and organizational intelligence is built through continuous accumulation.

AsiaInfo, an ICT and enterprise digitalization company with more than 30 years of experience, deployed TRAE IDE as an enterprise AI coding solution across over 6,000 seats. The R&D team's AI coding contribution rate reached 32%, and overall SDLC efficiency improved by 15%. AsiaInfo's vice president Li Xiaoping said the company found a new path for enterprise-scale intelligent R&D via TRAE.

Before adoption, AsiaInfo faced long delivery cycles, complex requirements, multiple roles, and high quality demands. It evaluated five core aspects: security compliance, delivery cycle reduction, quality stability, organizational replication, and quantifiable value. The resulting solution combines enterprise knowledge engineering, a full-process intelligent agent matrix including demand analysis, coding, testing, review, and operations agents, and a closed-loop data feedback system. TRAE ensures code is not stored or uploaded, supports custom model integration, and offers IDE plugins, CLI, and SOLO access. It also provides metrics such as call volume, adoption rate, and AI contribution rate, plus account and token management dashboards.

Leveraging TRAE, AsiaInfo built 'Tianshu Yuanheng Token ERP', a token management system based on its precision billing capabilities, with a three-layer measurement system that makes token costs calculable, controllable, and auditable. It supports company-, department-, project-, and scenario-level ROI insights.

The partnership followed a path of pilot validation, metric alignment, capability completion, organizational replication, and continuous operation. After full deployment, code defect escape rate in core teams dropped by 69% to 71%, easing delivery pressure and reducing rework costs. AsiaInfo has entered the stage of enterprise AI token governance, making AI investments accountable.

These deployments show TRAE moving beyond individual developer assistance to become a manageable enterprise R&D platform.