We document a novel empirical finding: regions in the United States that were more exposed to automation before the Great Recession experienced faster employment recoveries following that recession. This finding can be understood using a simple model of firm heterogeneity with capital accumulation. Comparing economies with different degrees of automation intensity, we find that less automation-intensive economies recover slower relative to more automation intensive economies following a transitory negative TFP shock, as they have a relatively lower level of automation capital and tend to reduce automation capital investment more after the shock. Consequently, they benefit to a lesser degree from the complementarity between automation capital and labour, resulting in a slower recovery of employment.