What stops working
The product role spent two decades administering scarcity: grooming backlogs, rationing capacity, brokering estimates, and writing requirement documents whose real function was to queue work for a constrained engineering middle. That middle is collapsing, and the administration around it has no constraint left to manage.
What it becomes
The classic triad balanced desirability, viability, and feasibility across three specialists. As the machines absorb feasibility, the team tilts toward the other two legs — and they were always the judgment work. Product moves from deciding when things get built to deciding what is worth building and why, then machining that intent precisely enough that an agent cannot proceed confidently in the wrong direction. Vague intent is no longer a slow start; it is a fast wrong turn. Specification is the leverage, not the whole job: product still carries the market-facing judgment the machines cannot — sequencing under uncertainty, pricing and positioning, which bet to make and which to refuse. The skill that compounds is writing intent; the accountability that remains is choosing what is worth intending.
The strongest objection
That this just renames "good product management" and changes nothing. Contested: the polarity of specification has flipped from overhead to primary artifact, and the role that owned the overhead now owns the most leveraged surface in the engine.