This research shows that production networks make automation‘s wage effects state dependent: the speed of wage recovery differs sharply across economies, with differences particularly pronounced during recessions. I embed a standard ordered-task automation block in an input-output economy and decompose aggregate wage dynamics into a recovery push from improving task fundamentals and a general-equilibrium drag from recessionary reweighting of network pass-through. A key implication is convex amplification: the recession differential in automation‘s wage effect becomes more negative as network tightness rises. Using robot exposure (IFR), wages (EU KLEMS), and annual WIOD input-output tables for 24 EU countries (2000-2014), with an IV based on peer-country adoption, I find that a one-standard-deviation increase in network tightness raises the recession wage loss from a one-standard-deviation robot exposure increase by about 0.11 log points. The framework also delivers two diagnostics summarizing when task-level recovery survives network propagation.