CyberAgent Accelerates with ChatGPT Enterprise and Codex

CyberAgent, a prominent Japanese internet company, is rapidly transforming its operations and enhancing its competitive edge through strategic adoption of OpenAI's cutting-edge AI tools: ChatGPT Enterprise and Codex. Across its diverse businesses, including internet advertising, media, entertainment, and gaming, CyberAgent leverages these platforms to accelerate workflows, elevate quality standards, and refine decision-making processes. This foundational commitment to AI has resulted in a remarkable 93% monthly active usage rate for ChatGPT Enterprise, underscoring a cultural shift towards intelligent, AI-powered productivity.

Secure AI Adoption with ChatGPT Enterprise

Even prior to the generative AI boom, CyberAgent had a strong foundation in AI research and development, particularly in digital marketing, marked by the establishment of "AI Lab" in 2016 and the introduction of "Kiwami Prediction AI" in 2020. However, the widespread emergence of generative AI, epitomized by ChatGPT's release in 2022, amplified the need for robust security and clear governance around AI tool usage.

The primary challenge was a pervasive uncertainty among employees regarding what information could be safely entered into AI tools, leading to hesitation and inconsistent adoption. To mitigate these risks and foster confident AI integration, CyberAgent strategically adopted ChatGPT Enterprise. This enterprise-grade solution offers tailored management features, allowing the company to configure data handling in accordance with specific requirements, bolstered by enterprise-level security and access controls. CyberAgent further cemented this secure environment by establishing internal guidelines for handling confidential information.

Ken Takao, Manager in the Data Technology Department of CyberAgent’s Corporate IT Promotion Division, highlights the impact:

"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."

This proactive approach has demystified AI usage, creating a secure and trust-filled environment where employees can confidently integrate AI into their daily tasks, such as research, drafting, and organizing key points, while ensuring human oversight retains final decision-making authority.

Cultivating AI Adoption: Beyond Mandates

What makes CyberAgent's AI integration story particularly compelling is its organic growth. The company does not impose specific tools but instead empowers teams, departments, and subsidiaries to evaluate and adopt solutions based on their unique objectives. Despite this non-mandate policy, ChatGPT Enterprise's monthly active usage soared to 93%.

This impressive adoption rate is a testament to CyberAgent's deliberate culture-building and robust training support from OpenAI. The company implemented several mechanisms to encourage proactive use:

Knowledge Sharing: Cultivating an environment for sharing successful prompts and use cases.
Internal Rankings: Providing employees with personalized AI tool usage data (non-evaluative) to promote self-awareness and engagement.
Proactive Engagement: A unique system where Slack bots follow up with employees who haven't used the tool recently, inquiring about challenges or alternative AI tools, and offering tailored suggestions.

Training sessions and workshops, co-designed with OpenAI, played a crucial role. From "ChatGPT Enterprise 101" to advanced custom GPT workshops and hands-on Codex sessions, these learning opportunities were tailored to different roles and levels of expertise. The high participation, often exceeding 100 employees per session, demonstrates a genuine eagerness to embrace AI skills. This comprehensive support framework has enabled employees to build practical, successful experiences with AI, moving beyond mere exposure to deep integration.

Codex: Elevating Design, Development, and Decisions

Codex, OpenAI's advanced code generation model, is rapidly gaining traction across CyberAgent, extending its utility far beyond basic coding tasks. Teams are leveraging Codex for critical upstream work, including design, alignment, and evaluation, where early and better decisions significantly reduce rework later in the development cycle. This strategic application of Codex contributes to a more efficient and higher-quality development pipeline.

Ken Takao identifies three pivotal use cases for Codex:

Design Evaluation: Reviewing, evaluating, and stress-testing design proposals from multiple perspectives.
Code Review Enhancement: Generating intelligent improvement suggestions during code review and facilitating the selection of optimal solutions.
Knowledge Documentation: Building and maintaining robust knowledge documents, such as AGENTS.md, to provide richer context for AI agents.

These applications translate into tangible benefits: enhanced design quality, faster alignment among stakeholders before implementation, and clearer rationale behind proposals, accelerating overall decision-making.

Tool	Primary Use Cases	Key Benefits
ChatGPT Enterprise	Research, drafting, organizing key points, general business operations, internal guidelines, secure data handling.	Secure data handling, enterprise-grade security, clear governance, increased employee confidence, widespread adoption, streamlined daily tasks.
Codex	Design reviews, code review suggestions, documentation, writing specifications, creating mockups, internal usage ranking system, game development.	Improved design quality, faster alignment, clearer rationale, enhanced development speed, reduced rework, boosted confidence, broader application in product development.

The impact of Codex is also spreading beyond traditional engineering roles. Non-developers are beginning to harness its capabilities for writing specifications, crafting mockups, and structuring work that interfaces with product and development teams. In a demonstration of its versatility, CyberAgent even utilized Codex to build its internal usage ranking system, further fostering engagement and visibility of AI adoption. For best practices in utilizing such models, teams often refer to resources on prompt engineering.

Real-World Impact: Speed, Quality, and Confidence

Across CyberAgent's development teams, Codex is increasingly valued not just as a tool for speed, but as an enabler for superior judgment and output quality.

Sou Yoshihara, a leading Codex user in CyberAgent’s AI Business Division, integrates Codex through MCP in Cursor for the design and implementation planning of Kiwami Prediction AI. Yoshihara notes, "Compared with other coding models, Codex gives the impression of producing higher quality proposals," highlighting its role in optimizing the entire development process.

This pattern is echoed in game development. Hidekazu Hora of GOODROID utilized Codex in the creation of the mobile puzzle game WormEscape, which reached soft launch in approximately one month. Hora emphasizes how Codex streamlined the flow from design to implementation to validation, providing critical support throughout.

"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." —Hidekazu Hora, Developer, GOODROID

Developers across the company report a common theme: Codex's usability and high-quality output are key drivers of its continued adoption. It not only accelerates work but also instills greater confidence in the development process, empowering teams to tackle new challenges and deliver higher-quality solutions. Managing and evaluating the performance of such agents is critical, as discussed in guides like evaluating AI agents for production.

The Future of AI Integration at CyberAgent

CyberAgent steadfastly regards AI as more than a fleeting trend; it is a fundamental turning point poised to redefine the internet industry's standards. With its "AI Lab" serving as a long-term engine for technical advancement and practical business application, the company is committed to continually expanding how AI creates value throughout the organization.

The unique aspect of CyberAgent's approach lies in its organic, bottom-up adoption model. Without blanket mandates, teams are empowered to choose the tools that best fit their needs, invest in learning how to utilize them effectively, and build internal systems that further propagate AI adoption. ChatGPT Enterprise and Codex have evolved beyond mere productivity tools; together, they are instrumental in CyberAgent's journey from merely introducing AI to embedding it deeply into everyday work, fundamentally redesigning operational methodologies. This strategic integration ensures that AI is not just a technology but a core driver of innovation and efficiency, setting a precedent for enterprise-level AI adoption. Moreover, ensuring security in such pervasive agentic environments is crucial, and insights from articles like designing agents to resist prompt injection are invaluable.

Original source

https://openai.com/index/cyberagent/

Frequently Asked Questions

How has CyberAgent integrated AI into its core business strategy?

CyberAgent views AI as a foundational technology supporting both business growth and operational design, not merely a collection of advanced initiatives. This perspective has driven significant investment, including the establishment of 'AI Lab' in 2016 for broad AI research and development, particularly in digital marketing. In 2023, the 'AI Operations Office' was launched to create a structured framework for leveraging AI to transform business operations. This strategic integration is centered on the widespread adoption of ChatGPT Enterprise and Codex, enabling rapid iteration through hypothesis testing, improving quality, and boosting productivity across its diverse business segments like advertising, media, and gaming.

What challenges did CyberAgent face before adopting ChatGPT Enterprise, and how were they resolved?

Before adopting ChatGPT Enterprise, CyberAgent faced significant challenges related to security, governance, and consistent AI usage. Employees were hesitant to use AI tools due to uncertainty about handling confidential information, leading to varied adoption rates and difficulties in knowledge sharing. To address this, CyberAgent implemented ChatGPT Enterprise, which provided robust management features, enterprise-grade security, and access control. This allowed the company to manage input data according to specific requirements, establish internal guidelines for confidential information, and create an environment where employees could confidently and securely incorporate AI into their daily tasks, ultimately expanding AI use across the entire organization.

How did CyberAgent achieve a 93% monthly active usage rate for ChatGPT Enterprise without mandating its use?

CyberAgent achieved a remarkable 93% monthly active usage rate for ChatGPT Enterprise through a deliberate strategy of culture building, continuous training, and proactive engagement, rather than enforcing mandates. The company fostered an environment where teams evaluated and adopted tools based on their specific objectives. Key initiatives included sharing knowledge and successful use cases, implementing internal rankings (for individual visibility only) to encourage engagement, and using Slack bots for 'light follow-ups' to understand usage patterns and offer tailored suggestions. Additionally, a series of training sessions and workshops, ranging from beginner introductions to hands-on custom GPT and Codex sessions, co-designed with OpenAI, significantly contributed to widespread adoption by empowering employees with practical AI skills.

Beyond code generation, what specific benefits and use cases has CyberAgent found for Codex?

CyberAgent has discovered that Codex offers substantial benefits beyond traditional code generation, particularly in upstream development phases. It's used to review, evaluate, and pressure-test design proposals from multiple perspectives, leading to better design quality. Codex assists in generating improvement suggestions during code reviews and selecting optimal options, fostering faster alignment before implementation and reducing rework. Furthermore, it aids in building and maintaining critical knowledge documents like AGENTS.md, providing richer context for AI agents. Non-developer roles are also leveraging Codex for tasks like writing specifications and creating mockups, extending its impact across product and development workflows by improving decision-making quality and speed.

Can you provide examples of how Codex has improved confidence and accelerated development at CyberAgent?

Codex has significantly boosted both confidence and development speed at CyberAgent. Sou Yoshihara, a top Codex user, praises its high-quality proposals for design and implementation planning, optimizing the entire development process. In game development, Hidekazu Hora from GOODROID utilized Codex for the game WormEscape, enabling its soft launch in approximately one month. Hora emphasized that Codex acted as a 'reliable partner' throughout the entire process—from discussing implementation to execution and validation. This partnership allowed him to quickly overcome roadblocks in unfamiliar areas, ensuring both quality and speed. This pattern of enhanced judgment and accelerated progress, even in areas without prior experience, is a common theme across CyberAgent's development teams.

What is CyberAgent's long-term vision for AI integration, and what makes their approach unique?

CyberAgent views AI not as a transient trend but as a transformative turning point for the internet industry, poised to become the next standard. Their long-term vision involves continuously expanding how AI creates value across the organization, driven by the 'AI Lab' for technical advancement and practical application. What makes their approach unique is the organic, non-mandated adoption model. Instead of blanket rules, teams select tools that align with their needs, learn through comprehensive training, and build internal systems (like usage ranking systems built with Codex) that naturally propagate AI adoption. This bottom-up integration shifts AI from mere introduction to being deeply embedded in daily work, fundamentally redesigning operational processes.

Stay Updated

Get the latest AI news delivered to your inbox.