Rakuten Doubles Speed Fixing Issues Using OpenAI Codex

Rakuten Halves MTTR Using OpenAI's Codex Coding Agent

Key Facts

What: Rakuten deployed OpenAI's Codex coding agent, reducing mean time to resolution (MTTR) by 50% while automating CI/CD reviews and enabling full-stack builds in weeks.
Who: Rakuten, a global technology and e-commerce conglomerate, working with OpenAI.
How: Codex acts as an always-on junior developer that writes features end-to-end, debugs bugs, traces failures, and creates pull requests.
Context: Part of broader enterprise adoption; companies including Cisco, Nvidia, Ramp, and Harvey have also rolled out Codex.
Related Development: OpenAI recently introduced Codex Security, an evolution of its Aardvark security research agent that scans code, validates vulnerabilities, and proposes fixes.

Lead paragraph

Rakuten has significantly accelerated its software delivery and incident response by deploying OpenAI’s Codex coding agent, cutting mean time to resolution in half, the companies announced. The Japanese tech giant now uses the AI coding agent to automate code reviews in CI/CD pipelines and complete full-stack builds in a matter of weeks rather than months. The deployment underscores OpenAI’s push into enterprise AI agents, with Codex helping development teams ship software faster and safer.

How Codex Transforms Rakuten’s Engineering Workflow

According to the official OpenAI announcement, Rakuten integrated Codex across its developer teams to handle repetitive and complex coding tasks. The agent can take a high-level description of a desired feature, write the necessary code, run tests, and submit a pull request without constant human supervision. When bugs arise, engineers simply paste the error message; Codex traces the failure, identifies the root cause, and patches the issue.

This capability has produced measurable results. Rakuten reduced its MTTR — the average time to fix problems after they are detected — by 50%. The improvement is critical for a company that operates large-scale e-commerce platforms, digital services, and financial technology products where downtime directly affects millions of users.

Codex also automates code reviews within Rakuten’s continuous integration and continuous deployment (CI/CD) pipelines. Previously time-consuming manual reviews are now accelerated, allowing human engineers to focus on higher-level architecture and innovation rather than routine checks.

Broader Enterprise Momentum for Codex

Rakuten is not alone in adopting the technology. OpenAI reports that usage of Codex, measured by weekly active usage, has grown fivefold. Major organizations including Cisco, Nvidia, Ramp, and Harvey have rolled out the agent across their developer teams, according to a Fortune report.

The tool positions itself as “a junior developer who never sleeps,” according to user guides and OpenAI’s positioning. Its ability to work end-to-end — from feature specification to tested and reviewed code — represents a step toward autonomous AI software engineering agents.

Codex Security: OpenAI’s Push Into AI-Native Application Security

The Rakuten deployment comes as OpenAI expands the Codex family with a dedicated security offering. Codex Security evolved from Aardvark, a security research agent that OpenAI began testing last year with a small group of customers. The platform analyzes entire code repositories, pressure-tests suspected vulnerabilities in sandboxed environments, generates proof-of-concept exploits to confirm real impact, and proposes fixes.

According to reports, the latest iteration leverages the reasoning capabilities of OpenAI’s frontier models combined with automated validation to reduce false positives. In one documented case, Codex Security scanned 1.2 million commits and identified 10,561 high-severity issues, demonstrating the scale at which these agents can operate.

This move places OpenAI in an increasingly competitive market for AI-enabled code security tools, competing with both traditional application security vendors and other AI laboratories developing similar capabilities.

Technical Capabilities and Enterprise Integration

Codex builds on OpenAI’s advanced reasoning models to understand context across large codebases. It can reason through complex dependencies, understand business logic, and generate production-ready code that adheres to an organization’s coding standards and security policies.

For Rakuten, the speed gains are particularly meaningful given the company’s diverse technology portfolio. Rakuten operates in e-commerce, fintech, telecommunications, and media, requiring coordination across multiple technology stacks and regulatory environments. Automating routine coding and review tasks allows its engineering organization to allocate talent toward strategic initiatives rather than maintenance work.

The agent’s ability to deliver full-stack builds in weeks — a process that traditionally could take months of coordinated effort across frontend, backend, and infrastructure teams — suggests meaningful productivity improvements at the organizational level.

Impact on Developers, Security Teams, and the Industry

For developers, Codex represents a shift from writing every line of code to directing and reviewing AI-generated work. This changes the day-to-day experience of software engineering, potentially reducing burnout from repetitive tasks while increasing the pace of innovation.

Security teams benefit from Codex Security’s ability to proactively scan massive codebases and validate vulnerabilities with working exploits rather than relying solely on static analysis. The combination of reasoning models with sandboxed testing helps provide higher confidence in both the existence and severity of issues.

The broader industry is watching OpenAI’s expansion into AI agents closely. The fivefold growth in weekly usage indicates strong market demand for tools that can meaningfully augment human engineering teams rather than simply offering code completion suggestions.

However, challenges remain. Recent Reddit discussions have noted occasional performance issues with Codex, including wait times between tool calls and lack of clear progress indicators during complex operations, highlighting that the technology is still maturing.

What’s Next for Codex and AI Coding Agents

OpenAI is expected to continue iterating on Codex, with potential improvements in reliability, context window size, and integration with more development environments. The company’s focus on enterprise AI agents suggests Codex will be central to its strategy for selling AI solutions to large organizations.

For Rakuten, the next phase likely involves deeper integration across more teams and potentially expanding use cases beyond current MTTR and CI/CD improvements. The company may also explore combining Codex with its own internal AI initiatives given its history of investing in advanced technology.

As more enterprises adopt these agents, questions around code ownership, security of AI-generated code, and appropriate human oversight will become increasingly important. OpenAI’s emphasis on automated validation in Codex Security indicates awareness of these concerns.

The Rakuten case study provides early evidence that AI coding agents can deliver quantifiable improvements in key engineering metrics. If the fivefold usage growth continues, Codex and similar agents could fundamentally reshape how software is built and maintained at scale.

Rakuten fixes issues twice as fast with Codex

Sources

Original Source

Related Topics

Comments