Cisco says Codex is now part of its engineering teams
Cisco says OpenAI’s Codex is being used across its engineering workflows, including AI Defense, build optimization, defect repair, and framework migrations. The company said the tool helped cut some work from weeks or quarters to days or hours.
Cisco says OpenAI’s Codex has moved from a developer assistant to part of its production engineering workflow, helping the company build software faster across large, regulated codebases.
In a post published by OpenAI, Cisco said it used Codex on systems that include multiple repositories, C/C++-heavy code, and enterprise security and compliance controls. The company said the goal was not to use Codex for simple code completion, but to see whether an AI system could handle real engineering work inside a large organization.
That test has already affected Cisco’s AI Defense product, which the company describes as an end-to-end AI security solution that protects against safety and security risks introduced by AI. Cisco said Codex helped write the majority of AI Defense and nearly every new feature it is building. DJ Sampath, senior vice president and general manager of AI Software and Platform at Cisco, said features that once took several quarters to reach customers dropped to weeks.
Cisco also said it worked with OpenAI on the Daybreak initiative, which brings together OpenAI models, Codex, and security partners to accelerate cyber defense and secure software continuously. As part of that effort, Cisco said it has governed access to GPT-5.5-Cyber, a model intended for cyber defenders.
The company said Codex was also used to help build Defense Squad, an open-source tool that moved from idea to the developer community in under a week.
Cisco said the appeal of Codex came from its ability to act with more autonomy inside complex systems. According to the company, Codex can understand large interconnected repositories, work in complex languages, run CLI-based autonomous compile-test-fix loops, and operate inside existing review, security, and governance frameworks.
Cisco said engineers gave OpenAI feedback on how those capabilities behaved in production, helping shape workflow orchestration, security controls, and support for long-running tasks. The company said that collaboration helped make Codex more suitable for enterprise use.
The most concrete results came in several engineering workflows. Cisco said Codex analyzed build logs and dependency graphs across more than 15 interconnected repositories and found ways to reduce build times by about 20%. The company said that saved more than 1,500 engineering hours per month across global environments.
Cisco also used Codex-CLI for defect remediation on large C/C++ codebases through iterative, agentic execution. The company said work that once took weeks now takes hours, and defect resolution throughput improved by 10 to 15 times.
In another case, Cisco’s Splunk teams used Codex to help migrate multiple UIs from React 18 to React 19. Cisco said the tool handled most repetitive changes on its own, turning weeks of work into days while engineers focused on judgment-heavy decisions.
Ryan Brady, a principal engineer in Cisco’s Splunk group, said the biggest gains came when the team stopped treating Codex as a tool and started treating it as part of the team. He said the team uses Codex to generate and follow a plan document so reviewers can understand both the process and the code.
Cisco said its work with OpenAI also helped shape Codex’s roadmap for enterprise use, especially around compliance, long-running task management, and integration with development pipelines. The company said Codex is now used across multiple business units, and teams are beginning to ask not how much work something will take, but how long a Codex run will take.
