The fast emergence of intensive robotization in combination with artificial intelligence implies a reappearance of the debate about the effects of innovation on the labor market. Many empirical studies have explored this phenomenon at the micro level, especially since the surge of innovation surveys, which use worldwide standardized indicators at the firm level. Most empirical studies suggest a robust, positive labor effect generated by new products, while the impact of process innovations on employment seems to be ambiguous. This paper offers a meta-regression analysis to seek some logical explanations for the results reflected in studies that apply the model proposed by Harrison et al. Our meta-regression suggests that the effect of sales growth due to new products on employment seems to be homogeneous and positive by different types of sub-samples. However, the labor effect of process innovation on employment depends on different circumstances. Its magnitude seems to be more negative for developing countries, manufacturing sectors, and periods of crisis. On the other hand, the magnitude tends to be positive for samples with the methodological approach (using instrumental variables), control variables, and high-tech sectors. The exercise is repeated, splitting the sample between developing and developed countries.