Cisco Made Codex a Teammate, and Quarters of Work Turned Into Weeks!

Hey everyone, it's me, Shii-chan! Today I found a really exciting adoption story about how Cisco put OpenAI's Codex to work, and I can't wait to share it with you!

OpenAI News openai.com

What was announced?

This one comes from OpenAI's News. Cisco, which has built and run some of the world's most complex, mission-critical software for decades, didn't treat the Codex coding agent as just a handy developer tool. Instead, they wired it straight into their production engineering workflows.

And they didn't stop at a light trial. They exposed Codex to massive multi-repository systems, C/C++-heavy codebases, and the security, compliance, and governance requirements of a global enterprise. In the process, Codex grew from a "developer productivity tool" into an AI engineering teammate that can operate at enterprise scale.

Why it matters

We hear a lot about using AI for code completion or small assists, but the thing I want you to notice here is the scale. When a company as large as Cisco, running systems that can never go down, can hand Codex real, responsible work, that's a big step.

What made Codex compelling for Cisco wasn't surface-level automation but agency. Specifically, Codex was able to:

• Understand and reason across large, interconnected repositories
• Work fluently in complex languages
• Run real workflows through CLI-based, autonomous compile-test-fix loops
• Operate within existing review, security, and governance frameworks

What changes

The clearest example is AI Defense, Cisco's end-to-end AI security solution that protects against the safety and security risks introduced by AI. Codex wrote the majority of AI Defense and nearly every new feature Cisco is building.

Features that would have taken several quarters to get into customers' hands dropped to weeks.

That's from DJ Sampath (SVP/GM, AI Software and Platform at Cisco). Features that used to take several quarters to reach customers now ship in weeks. What a jump!

Dive Deep

The original piece is packed with concrete numbers about the heavy workflows where Codex made a difference, so let me walk you through them.

First, cross-repo build optimization. Codex analyzed build logs and dependency graphs across more than 15 interconnected repositories and found the inefficiencies. The result: about a 20% reduction in build times, plus more than 1,500 engineering hours saved per month across global environments.

Next, defect remediation at scale (CodeWatch). Using Codex-CLI, Cisco automated defect repair on large-scale C/C++ codebases with iterative, agentic execution. Work that used to take weeks now finishes in hours, delivering a 10-15x increase in defect-resolution throughput and freeing engineers to focus on design and validation.

Framework migrations are a highlight too. When Splunk teams needed to migrate multiple UIs from React 18 to 19, Codex handled the bulk of the repetitive changes on its own, compressing weeks of work into days.

The biggest gains came when we stopped thinking about Codex as a tool and started treating it as part of the team.

That's Ryan Brady, a Principal Engineer in Cisco's Splunk group. He says the biggest gains showed up once they treated Codex as a teammate, having it generate and follow a plan document so reviewers could understand both the process and the generated code.

Security is exciting here too. Cisco is one of the leading security organizations in OpenAI's Daybreak initiative, which brings together OpenAI models, Codex, and security partners to accelerate cyber defense. As part of it, Cisco has governed access to GPT-5.5-Cyber, a model for cyber defenders. Cisco also used Codex to build Defense Squad, an open-source tool that went from idea to the developer community in under a week.

Wrap-up

Let me recap today's key points:

• Cisco embedded Codex deep into production engineering workflows, not as a standalone tool
• Codex wrote the majority of AI Defense and its new features, cutting delivery from several quarters to weeks
• Build optimization brought about a 20% speedup and saved 1,500+ engineering hours a month, and CodeWatch delivered 10-15x defect-resolution throughput
• A React 18 to 19 migration was compressed into days, and Daybreak, GPT-5.5-Cyber, and Defense Squad strengthened security
• It all adds up to a repeatable adoption model: deep technical partnership, real workloads, and leadership alignment from day one

Codex has become a meaningful part of how we think about AI-assisted development and operations going forward.

I'll close with Brad Murphy, a VP leading Cisco's Splunk Engineering team. Going forward, Codex has become essential to how they think about AI-assisted development and operations. If you want to put AI agents to work at enterprise scale, or you wrestle with running huge codebases, this story is for you!