Comparative advantage is the classical trade principle that specialisation should follow relative cost, not raw ability. An actor that is worse at everything than a rival can still be the rational choice for a given task, if it’s relatively less bad at that task than at the alternatives. It’s why a country, or a worker, keeps a role even when someone else could technically do it better.
The principle is easy to assume away in AI discussions, where a model that outperforms a human at every measurable task is treated as making that human’s labour worthless. That reasoning skips a step. If the compute a model needs to run is scarce and carries a real price, the decision an organisation actually faces is not whether the model can do the task, it’s whether running the model there is worth more than running it somewhere else. Comparative advantage reappears through the price of the constrained input.
The practical consequence is that absolute capability and economic allocation are different questions. A model can be superior at everything a workforce does and still leave some of that work with people, because the price of compute, not the ceiling on capability, decides where the constrained resource actually gets spent.