When anyone on a product team can produce a design, a spec, or working code, the question stops being who can make things. It becomes who decides.
The pattern that is holding up under real work: functions keep decision rights in their own domain. Engineering decides whether code ships. Design decides whether a design ships. Product decides what gets built. Anyone can propose in any direction, and the proposals are often good. But a proposal routes through the owner of that decision, no matter who or what produced it.
That only works if contribution and decision are explicit in both directions. An engineer proposing a design is contributing, and the design owner decides. A designer proposing requirements is contributing, and product decides. The practical move is to standardize the expected artifacts of the process rather than policing who creates them. The artifact carries the standard. The domain owner carries the judgment.
One boundary matters more than the others. Machine output about users is projection, not evidence. It is genuinely useful for sharpening hypotheses and finding the questions worth asking. It is not a substitute for putting work in front of real people. Teams that blur that line are not doing faster research. They are doing none.
The question we have not settled: as roles blur, how do accountability and decision rights stay legible over time? We do not have an answer yet. Writing that down is what working notes are for.