Cisco has embedded OpenAI’s Codex into core engineering pipelines, positioning the model as an active participant in large-scale software development rather than a standalone assistant. The company reports using Codex to write most of its AI Defense product and nearly every new AI feature under development.
Integration into production workflows
Cisco exposed Codex to multi-repository systems and C/C++-heavy codebases while maintaining enterprise security, compliance and governance controls. Engineers used Codex to run autonomous CLI-based compile-test-fix loops, analyze build logs and reason across large, interconnected repositories.
Cisco worked with OpenAI to adapt Codex for long-running engineering tasks, workflow orchestration and tighter security controls. The collaboration also involved participation in OpenAI’s Daybreak initiative and governed access to GPT‑5.5‑Cyber for cyber defense scenarios.
Measured results and engineering feedback
Cisco reported several measurable outcomes after broad deployment of Codex. According to company figures, more than 95% of new AI features were written by Codex. Cross-repo analysis and dependency optimization produced about a 20% reduction in build times and saved more than 1,500 engineering hours per month across global environments.
For defect remediation, Cisco used Codex-CLI and an automated process called CodeWatch to perform iterative, agentic repairs on large C/C++ codebases. The company reported a 10–15× increase in defect resolution throughput compared with prior manual workflows.
Teams also applied Codex to framework migrations: when Splunk teams migrated UIs from React 18 to 19, Codex handled the bulk of repetitive changes, compressing several weeks of work into days, the company said.
Engineering leaders framed the change as cultural as well as technical. “The biggest gains came when we stopped thinking about Codex as a tool and started treating it as part of the team,” said Ryan Brady, a Principal Engineer in Cisco’s Splunk group. Ching Ho, a member of Cisco’s engineering leadership, described collaboration with OpenAI on production readiness as rewarding. Brad Murphy, a VP leading Cisco’s Splunk Engineering team, said Codex “has become a meaningful part of how we think about AI-assisted development and operations going forward.”
Codex is now used across multiple Cisco business units, and the company says its feedback helped shape enterprise features such as compliance controls and support for long-running tasks.
Original source: OpenAI News