In 2016, radiologists at Stanford tested an AI system that diagnosed pneumonia from chest X-rays with 95% accuracy — better than most human radiologists. The hospital deployed it. Two years later, the radiologists were measurably worse at their jobs than they had been before the tool arrived. They had not been replaced. They had been subtly deskilled, their judgement quietly outsourced to a model they no longer quite understood.
This is the shape of the integration problem. Not that the machines will take the work, but that we will hand it over without noticing, and find — too late — that the know-how has left the building. This chapter is a field manual for the opposite: conscious participation, mutual competence, coordination as a daily practice rather than a theory.