CyberAgent moves faster with ChatGPT Enterprise and Codex

CyberAgent 是一家日本互联网公司，业务涵盖互联网广告、媒体与知识产权（IP）以及游戏。公司以“打造代表 21 世纪的公司”为愿景，依靠技术与创意优势，在国内外创造新的价值。

在 CyberAgent ，AI 不被视为若干孤立的高级项目，而是支撑业务增长与运营设计的基础技术。公司在这一领域持续投入：2016 年设立了 AI Lab ，开展与数字营销相关的各类 AI 研发；2023 年又推出了 AI Operations Office ，构建将 AI 作为业务转型手段的组织框架。

这些努力以采用 ChatGPT Enterprise 和 Codex 为核心。鉴于公司在广告、媒体与游戏领域的业务特点， CyberAgent 的竞争优势在于同时提升质量与生产力，并通过 AI 快速进行假设验证与迭代。将 ChatGPT Enterprise 建成 AI 环境的基础后，公司以其强有力的安全与管理能力，让员工在日常工作中放心使用 AI；如今，用 ChatGPT 做调研、起草和要点整理已成为常态，而最终决策权仍由人掌控。与此同时， Codex 也在设计讨论、代码审查与文档工作上帮助团队提速。

ChatGPT Enterprise 如何促成安全的 AI 落地

早在生成式 AI 出现之前， CyberAgent 就在推进面向广告领域的 AI 研发。2020 年，公司推出了 Kiwami Prediction AI ，推动将 AI 融入广告创意生产流程。

2022 年 ChatGPT 发布后，员工在日常工作中使用 AI 的场景迅速扩大。随着业务化使用增加，对更强的安全、治理与明晰规则的需求也随之上升。一个主要挑战是，员工不确定哪些信息可以安全地输入 AI 工具，导致顾虑重重、推进放缓；而且不同部门与个人的使用差异很大，难以在公司范围内共享经验与进行统一管理。

为此， CyberAgent 选择了 ChatGPT Enterprise ，该产品提供面向企业的管理功能，允许公司根据合同、配置与运营策略管理输入数据的处理方式，并提供企业级的安全与访问控制，使员工可以更有把握地将 AI 纳入日常工作。此外，公司还制定了内部机密信息处理指南，确保使用过程安全可靠。

企业 IT 推进部数据技术部经理 Ken Takao 表示， ChatGPT Enterprise 帮助营造了让员工放心使用 AI 的环境。

“像账号管理与使用可视化这样的企业功能，使得在不涉及机密数据的前提下支持业务性信息成为可能。结果是公司范围内的 AI 使用被大幅拓展，许多员工现在已在日常工作中频繁使用它。” —— Ken Takao, CyberAgent

为何采用能在非强制下广泛扩散

CyberAgent 的常态不是自上而下强制推行统一工具， ChatGPT Enterprise 亦如此：各团队、部门与子公司会在与其他工具比较后，根据自身目标决定是否采用。即便如此，该工具如今已覆盖几乎所有部门，月活跃用户率达 93%。

这一增长既源自公司内部有意营造的文化，也得益于 OpenAI 的持续培训支持。从文化建设看， CyberAgent 推行了鼓励主动使用的机制，例如分享 prompts 与成功案例，以及内部使用排名，帮助员工了解自己的 AI 工具使用情况（这些排名仅供个人查看，不用于绩效考核）。

公司还将跟进机制嵌入运营模型：当员工在一定期间未使用该工具时，内部机器人会在 Slack 上主动沟通，询问原因。若员工在使用其他 AI 工具，沟通有助于厘清工具用途与使用场景；若员工未使用任何 AI 工具，公司则会建议 AI 如何支持其具体工作，并提供实用技巧。

此外，由 OpenAI 主办的培训与工作坊也发挥了重要作用。尽管这些课程常在短时间内组织，仍有超过十场、每场逾百人参与，内容从面向初学者的“ ChatGPT Enterprise 101”到定制 GPT、 Codex 的实操课与内部黑客松不等。 CyberAgent 与 OpenAI 针对不同角色与成熟度设计学习路径，帮助员工以务实方式获得首批成功体验。

Codex 如何在落地前提升决策质量

Codex 在 CyberAgent 内的采用速度很快。团队不仅在代码生成上使用它，更多地把它用于上游工作，如设计、对齐与评估——这些环节若能早期做出更好的判断，能显著减少后期返工。

Takao 指出三个典型用途：

• 从多角度审视、评估并检验设计提案；
• 在代码审查中生成改进建议并在多个方案中做出选择；
• 构建并维护如 AGENTS.md 之类的知识文档，让 agent 在更丰富的上下文中运作。

他总结了使用 Codex 的三大收益：一是通过可从多角度评估的提案提升设计质量；二是在实施前更快达成对齐，从而减少后续返工；三是更清晰的提案理由帮助更快做出决策。

Codex 的应用也开始扩展到工程以外的岗位，非开发者用它撰写规格、制作原型、组织产品与开发相邻的工作；在某些情况下， Codex 甚至被用来构建支撑公司内部 ChatGPT 采用的系统，例如公司用它开发了内部使用排名系统，使得采纳情况更可见、更具互动性。

Codex 如何让团队更自信地加速推进

在 CyberAgent 的开发团队中， Codex 越来越被视为不仅能提速，更能提升判断质量的工具。AI 业务部的高频用户 Sou Yoshihara 在 Cursor 中通过 MCP 使用 Codex ，用于 Kiwami Prediction AI 的设计与实现规划。Yoshihara 认为，“与其他编码模型相比， Codex 的提案质量不给人感觉是粗糙的”，并把它视为优化整个开发流程的手段。

在游戏开发领域同样可见类似效应。GOODROID 的 Hidekazu Hora 强调，与 Codex 合作让从设计到实现再到验证的流程更容易掌控。Hora 在开发手游 WormEscape 时使用了 Codex ，该作在大约一个月开发后就进入了软启动。

“它像一个可靠的伙伴，从讨论实现到执行全程支持，帮助提升开发速度。即便在我没有经验的领域， Codex 也能更快地化解因知识欠缺带来的阻塞，我觉得它帮助我们在保证质量的同时也提高了速度。” —— Hidekazu Hora, GOODROID

在公司内部，开发者以不同方式使用 Codex ，但共同点是：易用性与输出质量是促使团队持续回归的原因。 Codex 不仅让人工作更快，也让人更有信心推进。

未来走向

CyberAgent 将 AI 视为非一时的热潮，而是互联网行业下一轮标准化的分水岭。以 AI Lab 作为长期的技术推进与业务落地引擎，公司计划继续扩大 AI 在组织内部创造价值的范围。

令人关注的不仅是采用规模，更在于采用的形态：在缺乏强制性统一规定的文化里，团队自主选择适合的工具、学习高效使用方法，并建立有助于扩散的制度。 ChatGPT Enterprise 与 Codex 已超越单纯的生产力工具，正共同推动 CyberAgent 从“引入 AI”迈向“将 AI 嵌入日常工作”，并由采纳走向对工作方式的重塑。

CyberAgent is a Japanese internet company engaged in businesses such as internet advertising, media & IP, and gaming. Guided by its vision of “creating a company that represents the 21st century,” the company leverages its strengths in technology and creativity to generate new value both domestically and internationally.

At CyberAgent, AI is positioned not as a set of limited advanced initiatives, but as a foundational technology that supports both business growth and operational design. The company has made continuous investments in this area. In 2016, it established “AI Lab” to conduct research and development of a wide range of AI technologies related to digital marketing. In 2023, it further launched the “AI Operations Office” to build an organizational framework for leveraging AI as a means of transforming business operations.

These efforts are centered on the adoption of ChatGPT Enterprise and Codex. Given the nature of its businesses in advertising, media, and gaming, CyberAgent’s competitive advantage lies in improving both quality and productivity at the same time, while also using AI to rapidly iterate through hypothesis testing. By making ChatGPT Enterprise the foundation of its AI environment, the company has enabled employees to use AI confidently in their daily work, supported by strong security and management capabilities. Today, using ChatGPT for tasks such as research, drafting, and organizing key points has become standard practice, while humans continue to hold final decision making responsibility. Codex is also helping people move faster in design discussions, code review, and documentation.

How ChatGPT Enterprise enabled secure AI adoption

Even before the emergence of generative AI, CyberAgent had been advancing research and development aimed at leveraging AI in the advertising domain. In 2020, the company introduced “Kiwami Prediction AI,” promoting initiatives to integrate AI into the advertising creative production process.

After ChatGPT was released in 2022, the use of AI in employees’ daily work expanded rapidly. As business use grew, the need for stronger security, governance, and clear rules became more important. A major challenge was uncertainty about what information could be safely entered into AI tools, which created hesitation and slowed adoption. Usage also varied across departments and individuals, making it difficult to share knowledge and manage AI use across the company.

To address those challenges, CyberAgent adopted ChatGPT Enterprise, which offers management features designed for enterprise use. ChatGPT Enterprise enables companies to manage how input data is handled in accordance with their specific requirements, including contracts, configurations, and operational policies. It also provides enterprise-grade security and access control features, allowing employees to confidently incorporate AI into their daily work. Furthermore, CyberAgent has established internal guidelines for handling confidential information to ensure safe and secure use.

Ken Takao, Manager in the Data Technology Department of the Corporate IT Promotion Division, says ChatGPT Enterprise helped create an environment where employees could use AI without hesitation. 

“With enterprise features such as account management and visibility into usage, ChatGPT Enterprise made it possible to support business use of a wide range of information, excluding confidential data. As a result, the scope of AI use across the company has expanded, and many employees now use it in their daily work.”

Why adoption spread without being forced

CyberAgent does not typically enforce the use of specific tools across the organization. ChatGPT Enterprise is no exception, with each team, department, and subsidiary evaluating it alongside other tools and adopting it based on their own objectives. Even so, it is now used across nearly all departments, reaching a monthly active user rate of 93%.

That growth was driven by deliberate culture building inside the company, along with ongoing training support from OpenAI.

From a cultural perspective, CyberAgent has implemented mechanisms to encourage proactive use, such as sharing knowledge including prompts and successful use cases, as well as internal rankings that allow employees to understand their own AI tool usage. (These rankings are designed so that individuals can only view their own data and are not used for evaluation purposes.)

CyberAgent also built follow-ups into its operating model. When an employee has not used the tool for a certain period, their bot reaches out on Slack to understand why. If the employee is using another AI tool, the conversation helps clarify which tools are being used and for what purpose. If they are not using any AI tools, CyberAgent suggests ways AI could support their specific work.

As part of efforts to encourage adoption, a light follow up is conducted via Slack when a tool has not been used for a certain period. Employees are asked whether they are using other AI tools and what challenges they may be facing. If needed, they are also given suggestions and practical tips on how to use AI more effectively in their work.

In addition, training sessions and workshops hosted by OpenAI have played a significant role. Despite often being organized within just a few days of announcement, more than 100 employees participated in each of over ten sessions. Those sessions ranged from beginner friendly introductions such as “ChatGPT Enterprise 101” to workshops on custom GPTs, hands on Codex sessions, and internal hackathons. Together, CyberAgent and OpenAI designed learning opportunities for different roles and levels of maturity, helping employees build their first successful experiences with AI in practical ways.

How Codex is improving decisions before implementation

Codex adoption is growing quickly across CyberAgent. Teams are using it for far more than code generation, especially in upstream work such as design, alignment, and evaluation, where better decisions early on can reduce rework later.

Takao points to three use cases in particular:

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

He also highlights three main benefits from using Codex:

1. Better design quality through proposals that can be evaluated from multiple perspectives.
2. Faster alignment before implementation, helping reduce rework later.
3. Clearer rationale behind proposals helped people make decisions faster.

CyberAgent is also seeing Codex spread beyond engineering. Non-developer roles are starting to use it for writing specifications, creating mockups, and structuring work adjacent to product and development. In some cases, Codex is even helping build the systems that sustain ChatGPT adoption internally. For example, CyberAgent used Codex to build its internal usage ranking system, making adoption more visible and engaging.

How Codex is helping teams move faster with more confidence

Across CyberAgent’s development teams, Codex is increasingly seen not just as a speed tool, but as a way to improve the quality of judgment.

Sou Yoshihara, one of CyberAgent’s top Codex users in the AI Business Division, uses Codex through MCP in Cursor and uses it for the design and implementation planning of Kiwami Prediction AI. Yoshihara says, “Compared with other coding models, Codex gives the impression of producing higher quality proposals,” and sees it not simply as a tool, but as a way  to optimize the development process as a whole.

The same pattern is visible in game development as well. Hidekazu Hora of GOODROID emphasizes that working with Codex made the flow from design to implementation to validation easier to navigate. Hora used Codex in the development of the game WormEscape, which reached soft launch after about one month of development.

“It felt like a reliable partner that supported the entire process from discussing implementation to carrying it out, helping increase development speed. With Codex, even in areas where I had no prior experience, I was able to resolve roadblocks caused by lack of knowledge more quickly, and I feel it helps achieve both quality and speed.”

Across the company, developers are using Codex in different ways, but a common theme has emerged: usability and output quality are what keep bringing teams back. Codex is not only helping people work faster. It is helping them move forward with more confidence.

What’s next 

CyberAgent views AI not as a temporary boom, but as a turning point that will become part of the next standard for the internet industry. With AI Lab as a long term engine for technical advancement and practical business application, the company plans to continue expanding how AI creates value across the organization.

What stands out about CyberAgent’s approach is not only the scale of adoption, but how that adoption has taken shape. In a culture without blanket mandates, teams choose the tools that work for them, learn how to use them well, and build systems that help adoption spread further. ChatGPT Enterprise and Codex have become more than productivity tools. Together, they are helping CyberAgent move from introducing AI to embedding it in everyday work, and from adoption to redesigning how work gets done.

Generated by RSStT. The copyright belongs to the original author.

Source