CyberAgent's internal Claude Code rollout

CyberAgent says about 2,000 internal users use Claude Code as adoption expands from individual experimentation to organisation-wide practice. Its engineering teams discuss development-process automation, code review, security review, operations, permission management, token cost, training, and reusable organisational instructions. The company says its longer-term goal is to automate the development process by 2028, while retaining explicit decisions about work that people should continue to own.

The first-party engineering account describes governance and operating practices rather than a measured productivity result. CyberAgent says teams need visibility into usage, cost, permissions, configuration, and security, plus tests, builds, screenshots, logs, and CI that let agents verify their work. It does not publish a before/after productivity baseline, quality result, cost-benefit calculation, or causal evaluation. The approximately 2,000-user figure is rollout evidence, not an outcome.

This independently supports the scale and software-delivery scope of cyberagent-moves-faster-with-chatgpt-enterprise-and-codex, whose separate OpenAI customer story reports 93% monthly active ChatGPT Enterprise usage and qualitative workflow benefits. Neither source establishes a company-wide ROI.

Related

• gitlab-claude-enterprise-internal-pilot — a smaller internal Claude pilot with vendor-reported productivity claims and measurement limitations.
• atlassian-claude-rovo-code-review-outcome — observational software-delivery evidence with a defined repository population.
• anthropic — Claude's provider.

Source: CyberAgent Developers Blog | 2026-06-18 (Japanese)