This systematic review critically assesses the contemporary evidence on whether artificial intelligence (AI) implementation delivers productivity gains that consistently surpass those of previous general-purpose technologies (GPTs), such as electricity and information and communication technology (ICT).A structured literature search was conducted using the Consensus research platform, compliant with PRISMA guidelines. The search strategy involved 22 targeted queries across 7 thematic groups, identifying 1,100 papers. Following a multi-stage screening process, 50 highly relevant empirical studies, systematic reviews, and meta-analyses were selected for narrative synthesis. Data were extracted to analyze AI's productivity impact, its comparison to previous GPTs, and the heterogeneity of its effects.The evidence confirms that AI implementation delivers measurable and often substantial productivity gains at the firm and process levels across diverse sectors. Mechanisms include cost reduction, process automation, and innovation acceleration. However, the claim that AI consistently outperforms historical GPTs is not fully supported. The impact of AI echoes the "productivity paradox" observed with earlier technologies, where macroeconomic productivity growth remains modest and significant lags persist. Furthermore, the benefits are unevenly distributed, disproportionately favoring large, digitally mature firms.
Nima Taheri Hosseinkhani (Mon,) studied this question.