Global network technology giant Cisco has recently announced that by deploying the OpenAI subsidiary Codex model on a large scale, it has successfully transformed "AI-native" development into a core part of its enterprise software building process. This shift marks the role of AI within enterprises evolving from a simple assistant tool to an "engineering teammate" with practical capabilities.

image.png

Cisco did not limit Codex to code completion but deeply integrated it into complex production workflows. Facing large codebases involving complex languages like C/C++, Cisco used AI agents to achieve an automated "compile-test-fix" loop. This application of agentic AI reduced manual repair work that previously took weeks to just a few hours.

In terms of actual results, Cisco's CodeWatch system significantly improved defect resolution efficiency, with throughput increasing by 10 to 15 times. At the same time, through the optimization of cross-repository build processes by Codex, Cisco saved over 1,500 engineering hours per month globally, with build time reduced by about 20%. Even in complex UI framework migration tasks, work that originally took weeks was compressed into just a few days.

The Cisco engineering team pointed out that the real productivity leap came from a change in mindset: viewing AI as a team member. In the future, Cisco and OpenAI will continue to cooperate, exploring deeply in areas such as compliance, security controls, and long-term task management, jointly defining new standards for AI-native engineering.