This paper presents an aggregative model of an economy's production function that links the level and distribution of automation technology at the micro- (firm-) level to macro features including TFP, factor shares and the elasticity of substitution between labor and capital. To test implications of the theory for US industry-level labor shares, I develop a novel aggregate measure of tasks performed by workers in the US economy. I model the production process as a set of tasks that can be performed by labor or automation capital. Aggregating over firms that operate capital with differing degrees of automation, total output of the economy is given by a Constant Elasticity of Substitution (CES) function, but with parameters determined endogenously by the distribution of automation technology across firms. This model of the aggregate production function can reconcile three important empirical findings on US production and growth that the canonical CES model cannot: declining labor shares, aggregate capital-labor complementarity, and capital-biased technical progress. Using industry-level data, including a novel measure of aggregate task inputs into production, I find evidence that automation has been a significant driving force of US industry-level labor shares for the past several decades.