The Productivity Effects of Artificial Intelligence: A Comparative Analysis of a New General-Purpose Technology and its Transfer

Purpose: This study synthesizes and evaluates the empirical evidence on the transfer and diffusion of artificial intelligence (AI) by analyzing whether its implementation delivers productivity gains that consistently exceed those of previous general-purpose technologies (GPTs), such as information and communication technology (ICT) and electricity. It aims to clarify the magnitude, mechanisms, and contextual dependencies of AI's impact, framing the issue as a challenge in technology transfer from development to widespread economic application. Methodology: A systematic literature review was conducted following the PRISMA 2020 framework. The search utilized the Consensus academic search engine, covering sources like Semantic Scholar and PubMed, with 22 targeted queries across seven thematic groups. The process involved identifying 1,100 papers, screening 630, assessing 491 for eligibility, and conducting a full-text analysis and narrative synthesis of the 50 most relevant studies. Methodologies of the included papers range from large-scale panel data regressions and randomized controlled trials to systematic reviews and macroeconomic analyses. Findings: The evidence consistently shows that AI implementation delivers measurable productivity gains at the firm and process levels across various sectors. Key mechanisms for this value capture include cost reduction, process automation, skill-biased labor enhancement, and innovation acceleration. For instance, specific applications like generative AI have been shown to reduce task completion time by 40% and improve output quality by 18%. However, the evidence that these gains consistently surpass those of earlier GPTs is nuanced, revealing lags and barriers characteristic of historical technology transfers. The diffusion of benefits is uneven, disproportionately favoring larger, digitally mature firms with higher absorptive capacity. At the macroeconomic level, AI's contribution to aggregate productivity growth remains limited, echoing the "productivity paradox" observed during the initial transfer of ICT and electricity. Implications: The findings suggest that while AI is a potent productivity driver, realizing its full economic potential is contingent on overcoming key barriers to technology transfer, including the need for complementary investments, organizational restructuring, and workforce upskilling. For policymakers and technology managers, this underscores the need for strategic initiatives that address expertise gaps and integration challenges, thereby fostering more inclusive and widespread technology diffusion and productivity growth. The historical parallels with previous GPTs suggest that the transformative impact of AI may materialize over a longer time horizon than currently anticipated, dependent on the efficiency of these transfer mechanisms.