Rakuten Fixes Issues Twice as Fast With OpenAI Codex

Rakuten is fixing production issues twice as fast after deploying OpenAI's Codex coding agent across its engineering operations.

The e-commerce and fintech giant, with 30,000 employees worldwide, integrated Codex into its software delivery pipeline over the past year. The result: a roughly 50% reduction in mean time to recovery (MTTR), according to the company.

"We don't just care about generating code quickly," said Yusuke Kaji, General Manager of AI for Business at Rakuten. "We care about shipping safely. Speed without safety is not success."

Beyond incident response, Rakuten uses Codex in its CI/CD pipeline for automated code review and vulnerability checks. The system applies internal coding standards automatically, so teams can ship fast without cutting corners on security.

The impact extends to new development. One example: building a mobile app version of an existing web-based AI agent service. Codex implemented a full-stack solution with a Python/FastAPI backend and Swift/SwiftUI iOS app—including all backend APIs—without step-by-step human instruction. The project went from a quarter to weeks.

"The latest Codex models can read between the lines," Kaji said. "Even if the requirements are not perfectly defined, it understands what we're trying to build."

As Codex takes on more code generation work, Rakuten engineers are shifting to writing clearer specifications and verifying outputs against measurable standards.

"Our role is not to check every line of code anymore," Kaji said. "Our role is to define clearly what we want and establish how to verify it."