This post, which opened CyberAgent’s developer Advent Calendar for 2025, is written by the executive officer in charge of technology at CyberAgent. It announces the vision the company set out at an internal event in October 2025 — to aim for full automation of its development process, which it equates with AI maturity Level 4, by 2028 — and describes how the company intends to restructure its engineering organization around that aim. The author is explicit that the goal is not automation “to make things easier” but a challenge in which AI agents and engineers work together to reinvent how development is done.
Most of the post is a response to the questions engineers asked after the vision was presented, which the author groups into five kinds of anxiety, and to each of which the company attaches a concrete measure. It is a statement of policy and intent from the company’s own management, published on its own developers’ blog, rather than an account of results.
Key Points
- The five anxieties it lists are losing one’s career because one’s specialism or role is replaced by AI; skills deteriorating because relying on AI yields results without deep understanding, hollowing out junior engineers; unfair evaluation because contributions made with AI are hard to measure; engineers’ value shifting from how to build to what to build; and structural inequality between teams, since the benefit of generative AI varies with a product’s situation and environment.
- The author presents AI lowering the cost of “How” — implementation — as engineers’ biggest opportunity, and argues against treating AI as a subcontractor for tedious work or only as a way to cut effort; the stated goal is to treat AI as the strongest partner and to assume a world in which engineers and AI create together.
- The company’s engineer evaluation system, the JB Career Program, is to be fully renewed the following spring for the first time since it was drawn up in 2019. What stays constant is solid technical skill that guarantees high-quality output; what changes is the ability to create value and outcomes with AI as a weapon. The post’s reasoning is that the engineer remains responsible for the quality and reliability of the final output, and cannot evaluate or integrate AI’s work without strong fundamentals.
- Four career ladders are being drawn up: a renewed tech-lead engineer track that flattens previous job categories such as server-side and native so engineers can cross domains using AI; a new business-lead engineer role that goes into the business side and owns everything from problem discovery and requirements through design, implementation and operation; a strengthened specialist track, with 25 focus areas defined (3 in AI utilization, 12 in product quality, 10 in business expertise); and engineering managers, responsible among other things for building new AI-assisted development processes.
- Against skill deterioration the company says it will strengthen training for junior engineers, on the reasoning that the basic technical skill needed to evaluate and correct AI output rather than accept it uncritically becomes more important, not less.
- Against unfair evaluation it is starting an engineer-specific AI ranking which, unlike the company-wide version, is to evaluate primarily on engineering quality and technical maturity, making AI-assisted contributions visible in terms of technical depth and level of automation.
- Against inequality between teams it set up a dedicated AI Driven Promotion Office in August 2025 to support, across the company, an autonomous development structure in which engineers and AI agents work together — through reskilling programmes, infrastructure for visualizing and evaluating productivity, and events such as an inter-department contest in AI agent development skills called “AI Agent Arena”.
Context
The post places the change against continuing layoffs in the tech industry, which it links to AI replacing part of coding, the value of work that only builds what it is told to build collapsing, and a polarization of engineers that it expects to accelerate. It is written by the company’s own management about the company’s own plans, and describes measures announced or under way rather than their effects.