A new hybrid role blending product management and engineering is spreading across hiring pipelines, and companies say the hard part isn't finding people who can code but finding people who can decide what's worth building. The signal matters more than the job title.
The global logic is straightforward. When code generation becomes cheap and near-instant, the marginal value of an engineer who executes someone else's spec collapses toward the cost of the tool that does the same thing faster. What appreciates is everything upstream of the keyboard: framing the right problem, reading the domain, designing cheap reversible experiments, and killing your own pretty ideas when the metrics say so. This is also why org charts are flattening. Layers of middle management existed largely to translate strategy into tasks and route those tasks to executors. Compress the executor layer with AI and much of that coordination overhead loses its reason to exist. The winners in this shift are individual contributors who can hold a business goal and a code editor in the same head. The losers are structures built to slice work into tickets.
For Japan, this is not a soft skills memo. It is a direct challenge to the industry's operating system. The multi-layered subcontracting model, where a prime SIer wins the contract, writes the specification, and pushes implementation down through tiers of vendors, is the purest possible expression of the task-taker economy. Its entire margin structure assumes that turning fixed specs into working code is scarce and billable by the person-month. AI erodes exactly that assumption. When the build step is commoditized, the value migrates to whoever owns the requirements and the customer relationship, and the lower tiers face the sharpest squeeze.
RPA-heavy transformation programs face a parallel reckoning. Much of Japan's automation spend went into scripting brittle workflows on top of processes nobody redesigned. A product-minded engineer would question whether the process should exist at all, which is a harder and more valuable conversation than recording more clicks.
The practical move for Japanese enterprises and SIers is to stop measuring engineers by delivery velocity against specs and start rewarding those who reshape the spec. That means giving domestic dev teams direct line of sight to business KPIs, real authority to run small experiments, and permission to push back, the same pushback culture the article prizes and that rigid Japanese reporting hierarchies most suppress. Firms that make that cultural shift early convert AI from a cost threat into leverage. Those that keep optimizing the ticket queue are optimizing a role the market is quietly retiring.
