93% Active Users, Zero Mandates: How CyberAgent Scaled ChatGPT Enterprise and Codex

Hey everyone, it's me, Shii-chan! Today's story is about how a big Japanese internet company rolled AI out across the whole company, and honestly the way they did it is so thoughtful that I got excited reading it!

OpenAI News openai.com

What was announced?

OpenAI News published a customer story about CyberAgent, a Japanese internet company working in areas like internet advertising, media & IP, and gaming. Guided by its vision of "creating a company that represents the 21st century," CyberAgent is using ChatGPT Enterprise and Codex to move faster.

What stands out is that they treat AI not as a set of limited advanced projects, but as a foundational technology that supports business growth and operational design. They set up an "AI Lab" research group in 2016 and an "AI Operations Office" for transforming operations in 2023.

Why it matters

When generative AI spreads inside a company, people start worrying: what information is safe to put into these tools? That uncertainty slows adoption. And when usage varies wildly by team and person, it gets hard to share know-how or manage AI use at all.

CyberAgent hit exactly those walls. That is why getting security, governance, and clear rules right mattered so much. This story is full of hints for any large organization trying to move from using AI nervously to using it as a matter of course.

What changes

By building on ChatGPT Enterprise, which comes with management features for enterprise use, employees can confidently use a wide range of information (excluding confidential data) in their daily work. With account management and visibility into usage, the hesitation of "is it okay to paste this?" drops away.

And CyberAgent does not force specific tools on the whole company. Even so, ChatGPT Enterprise now reaches nearly every department, with a monthly active user rate of 93%. Spreading that far without any mandate is pretty impressive, right?

Dive Deep

The growth came from a combination of deliberate culture building and ongoing training support from OpenAI.

• They share prompts and successful use cases, and offer internal rankings where employees can see their own usage (each person can only view their own data, and it is never used for evaluation).
• When someone has not used the tool for a while, a bot reaches out on Slack to ask why. If they use another AI tool, the chat clarifies what for; if not, CyberAgent suggests ways AI could help their specific work.
• OpenAI-hosted training and workshops were a big push too. Even when organized within days, each of over ten sessions drew more than 100 participants, ranging from "ChatGPT Enterprise 101" to custom GPT workshops, hands-on Codex sessions, and internal hackathons.

The way Codex is used is the fun part. It shines not just in code generation but in upstream work like design, alignment, and evaluation. Ken Takao from the Corporate IT Promotion Division points to three use cases:

• Reviewing, evaluating, and pressure-testing design proposals from multiple perspectives.
• Generating improvement suggestions during code review and picking among options.
• Building and maintaining knowledge docs like AGENTS.md so agents can operate with richer context.

The benefits are threefold too: better design quality, faster alignment before implementation (which cuts rework), and clearer rationale that speeds up decisions. Codex is spreading beyond engineering into spec writing and mockups, and CyberAgent even built its internal usage ranking system with Codex.

You also hear it from the field. Sou Yoshihara, a top Codex user in the AI Business Division, uses Codex through MCP in Cursor for the design and implementation planning of Kiwami Prediction AI.

"Compared with other coding models, Codex gives the impression of producing higher quality proposals."

Game development shows the same pattern. Hidekazu Hora of GOODROID used Codex to build the game WormEscape, which reached soft launch after about one month, and says the flow from design to implementation to validation got much easier.

Wrap-up

• CyberAgent embedded AI company-wide around ChatGPT Enterprise and Codex, reaching a 93% monthly active user rate.
• No mandates: adoption spread through culture building (rankings, Slack follow-ups) and OpenAI training.
• Codex shines in upstream work like design, review, and docs, cutting rework and speeding up decisions.
• Adoption reached beyond engineers into internal tooling and game development.

If you are the person tasked with rolling out AI in a big org, or an IT lead or manager fighting the "nobody actually uses it" problem, this story is for you!