Diverse impacts of AI investments on productivity gains: Effects of industry and innovation characteristics

As artificial intelligence (AI) becomes a key driver of innovation, its role in boosting productivity has gained growing attention. Yet most existing evidence links AI adoption to productivity at the occupation or national level; firm-level studies remain scarce. Using U.S. job-posting data for 2014–2023, we examine how AI investment relates to firm Total Factor Productivity (TFP). Our results show that AI investment raises firm-level TFP, but unevenly. The impact is industry-specific, reflecting the varying availability of the necessary complementary assets: Manufacturing captures the largest gains, while in services the benefits are concentrated in information, utilities & construction, education & health, and public administration. Productivity gains are not systematically related to a firm's AI-exposure score—a measure of the potential for its required job skills to be exposed to AI—although early adopters receive a slight additional boost. AI creates value through both exploration and exploitation, but the estimated productivity returns are stronger for exploitative deployments. These findings highlight the need for sector-specific, purpose-built AI strategies. By moving beyond the notion of AI as a universal productivity engine, our study underscores the importance of tailoring AI investment to the distinctive challenges and dynamics of each firm.