Rakuten Integration of OpenAI Codex for Engineering Efficiency
Rakuten has integrated OpenAI Codex into its engineering workflows to accelerate software delivery and incident response while maintaining security standards. This integration has resulted in a approximately 50% reduction in mean time to recovery (MTTR) and the ability to compress development cycles for complex projects from quarters to weeks.
Accelerating Incident Response and Reducing MTTR
Rakuten uses Codex to compress the time between system alerts and resolution. By integrating Codex with KQL (Azure’s query system for logs and telemetry), engineers can more efficiently monitor APIs, analyze signals, and identify root causes.
Instead of manually synthesizing queries, logs, and patches, site reliability engineering (SRE) teams use Codex to suggest fixes and accelerate remediation. This approach has reduced the mean time to recovery (MTTR) by approximately 50%, allowing Rakuten to fix production issues twice as fast.
Automating Safety and Security in CI/CD
To prevent deployment bottlenecks, Rakuten integrates Codex directly into its CI/CD pipeline to automate code reviews and vulnerability checks. The system ensures that all code changes align with company-specific expectations by feeding internal coding principles and standards into the Codex workflows.
As Yusuke Kaji, General Manager of AI for Business at Rakuten, states:
‘We provide our internal coding principles to Codex. Using the same principles, it reviews whether the code aligns with our standards.’
This automation ensures that safety checks are performed consistently without lowering engineering standards as shipping speed increases.
Scaling Autonomous Full-Stack Development
Rakuten employs Codex to execute large, ambiguous projects from initial specifications to working implementations, reducing the dependency on perfectly defined requirements. This "AI-nization" strategy allows the model to interpret partial requirements and produce usable artifacts autonomously.
In one specific instance, Rakuten used Codex to build a mobile app version of an existing web-based AI agent service. Codex performed the full-stack implementation, which included:
- A Python/FastAPI backend
- A Swift/SwiftUI iOS application
- All necessary backend APIs
This project was completed without step-by-step human instruction, reducing the development timeline from one quarter to several weeks.
Evolution of the Engineering Role
With the adoption of Codex, Rakuten is shifting the primary responsibility of its engineers from manual code review to specification authorship and verification. The focus has moved toward defining clear objectives and establishing measurable standards to verify AI-generated outputs.
To facilitate this transition, Rakuten has implemented hands-on workshops across engineering, product, and non-technical teams to scale the use of autonomous development across the organization.