Rakuten fixes issues twice as fast with Codex

Source: OpenAI Date: 2026-03-11 URL: https://openai.com/index/rakuten

Summary

Title-only: A Codex case study featuring Rakuten — the Japanese e-commerce and technology conglomerate — achieving a 2x speedup in software issue resolution using Codex. This is a quantified productivity claim: Codex is not just helping developers write code but measurably reducing time-to-resolution for software bugs and issues at Rakuten’s engineering scale. March 2026 puts this after Codex GA (October 2025) and deep into Rakuten’s enterprise deployment.

Implications

The productivity measurement thread. “Twice as fast” is a concrete, auditable claim — it implies Rakuten measured issue resolution time before and after Codex deployment at scale. Quantified productivity claims are the most persuasive enterprise sales evidence. If Rakuten publishes this number, other enterprise engineering organizations with similar issue volumes (tens of thousands of engineers) will benchmark their Codex deployments against it.

Japan engineering efficiency. Rakuten is one of Japan’s largest technology employers — a productivity doubling in issue resolution has meaningful economic impact at their scale. The case study is also evidence that Codex is reliable enough for production use in Japanese enterprise engineering (Rakuten’s stack includes diverse technologies) — addressing the concern that AI coding agents are only effective for greenfield English-language codebases.