The Engineer AI Banzuke (エンジニア版AI番付) is an assessment that CyberAgent designed to make visible how far each of its development organizations has taken AI adoption, as a development-specific counterpart to the company-wide AI Banzuke. Rather than asking only whether an organization uses AI, it scores maturity on two axes — “engineering quality and technical maturity” (scope of AI use, depth of automation, division of roles between humans and AI, integration into the development workflow, results, incident response and AI infrastructure) and “promotion structure and technical culture” (level of use within the organization, tooling and guidelines, knowledge sharing, promotion structure, measurement of effects, communication to the rest of the company, and alignment with goals and strategy) — across fourteen evaluation items, and expresses the result as one of seven ranks named after the ranks of a sumo banzuke.
Usage
The seven ranks run from makushita, where AI use stays at the level of individuals, through jūryō (starting to copy other organizations’ examples), maegashira (partially adapting AI to the organization), komusubi (comprehensively adapting it), sekiwake (organizational success cases established) and ōzeki (standardization, measurement and outward communication under way), up to yokozuna, an AI-first development organization. The first round covered 44 development organizations. Each organization reported its own status, submitted evidence supporting the report, and the operators checked that evidence against the evaluation items, with a deliberation council reviewing cases where needed before ranks were finalised. A second round was planned for announcement in October 2026.
When It Applies
The operators present it as both an evaluation scheme and a diagnostic tool for development organizations moving to AI-first ways of working, in support of CyberAgent’s stated aim of automating the development processes across the company by 2028. They chose two axes because they consider both technically advanced use and the organizational capacity to spread it reproducibly necessary: a few engineers with advanced practices that are not shared, or study sessions without AI in the actual development process, both leave room to grow. Self-reporting was chosen so that working out one’s own position would itself prompt thinking about AI strategy, while evidence and deliberation were added because self-reports alone cannot ensure fairness. The operators acknowledge that a ranking format risks the result taking on a life of its own, and say they consistently emphasised clarifying what each organization should improve next rather than ranking it. All of this is the operators’ own account of a single company’s first round; the source reports observed tendencies rather than measured outcomes.