OpenAI says CyberAgent is using ChatGPT Enterprise across nearly all departments

CyberAgent has moved generative AI from isolated experiments into everyday work across much of the company, according to a new OpenAI case study published on April 9, 2026. The Japanese internet group said ChatGPT Enterprise has reached a 93% monthly active usage rate, while Codex is increasingly being used for design discussions, code review, documentation, and other upstream engineering tasks.

CyberAgent’s 93% usage rate points to a broad internal rollout

The scale of adoption is notable because CyberAgent says it does not force teams to use a single tool. Instead, departments and subsidiaries evaluate AI systems on their own, yet ChatGPT Enterprise has still spread across nearly all parts of the organization. OpenAI’s account describes the platform as the foundation of CyberAgent’s AI environment, with employees using it for research, drafting, and organizing key points while keeping final decision-making with people.

CyberAgent’s internal approach appears to combine enterprise controls with day-to-day encouragement. The company says it adopted ChatGPT Enterprise to reduce hesitation around what data could safely be entered into AI tools, and it has also set internal guidelines for handling confidential information. That matters operationally: broad AI use inside a large company depends less on novelty than on whether security, access control, and governance are clear enough for teams to rely on the system consistently.

Codex is being used before code is even written

The second part of the rollout is Codex, which CyberAgent says is helping teams move faster in design alignment, code review, and documentation. The case study says workers are using it not just to generate code, but to review proposals from multiple angles, compare implementation options, and build knowledge documents that give agents more context. That shifts generative AI from a downstream productivity aid into a tool for earlier technical judgment.

OpenAI also says adoption has spread beyond engineering, with non-developer roles using Codex for writing specifications, making mockups, and structuring adjacent product work. In one example cited by the company, CyberAgent used Codex to build an internal usage ranking system to make AI adoption more visible. The detail is small, but it shows how quickly generative AI can become part of the machinery used to manage its own rollout.

Why this matters for the enterprise AI market

For the generative AI market, CyberAgent is another sign that commercial value is being measured less by model benchmarks than by workflow penetration. A 93% monthly active usage rate suggests the limiting factor is no longer awareness of AI tools, but whether they can be embedded into specific company processes with enough trust, training, and governance to stick. That is especially relevant in advertising, media, and gaming, where speed, iteration, and quality control all matter at once.

OpenAI said the company’s internal adoption was supported by training sessions and workshops, including sessions with more than 100 participants at a time. The operational implication is straightforward: enterprise AI rollouts are becoming change-management programs, not just software deployments. CyberAgent’s case now reads as a practical benchmark for how generative AI moves from optional assistant to standard working layer inside a large internet business.