Introduction
As more companies experiment with artificial intelligence (AI) to enhance profitability, the debate surrounding its implications for workers intensifies. In the United States, a stark disconnect between soaring stock market gains and declining job offers (excluding the agricultural sector) has sparked media discussions about job destruction driven by technology.
The Pessimistic Narrative
Weekly news reports highlight companies leveraging AI for office tasks, typically performed by recent graduates or entry-level professionals. A Senate Committee report suggests that AI and automation could eliminate nearly 100 million jobs in the US over the next decade. Pessimistic voices even cite prominent economists predicting that AI’s impact on productivity growth will be modest, while its effect on employment will be clearly negative due to the automation of numerous tasks and jobs.
A More Nuanced View
We disagree with this bleak outlook. Our recent research indicates the situation is more complex and less severe than the pessimistic narrative suggests. Regarding productivity growth, AI can operate through two distinct channels: automating tasks in goods and service production and automating tasks in generating new ideas.
Erik Brynjolfsson and his co-authors found that AI generative models increased worker productivity by nearly 14% in the first month of use at a US software company, stabilizing around 25% after three months. Similar productivity gains have been observed in diverse knowledge workers, with the initial improvement being more pronounced for less productive workers, thus reducing inequality within companies.
Assessing AI’s Impact on Growth
In a 2024 article, two of us (Aghion and Bunel) estimated AI’s impact on growth potential over the next decade using two alternative methods. The first method capitalized on similarities between AI revolution and past technological revolutions, while the second employed Daron Acemoglu’s task-based model combined with empirical study data.
Using the first method, we estimated that the AI revolution could add between 0.8 and 1.3 percentage points to annual aggregated productivity growth over the next decade. The Acemoglu task-based model, along with our interpretation of recent empirical literature, estimated an increase between 0.07 and 1.24 percentage points annually, with a median of 0.68 (Acemoglu forecasts only a 0.07 percentage point increase).
Our median estimate should be viewed as a lower bound, as it doesn’t account for AI’s potential to automate idea production. Moreover, our estimates don’t consider possible growth obstacles, such as the lack of competition in various AI value chain segments already controlled by leading digital revolution firms.
AI’s Impact on Employment
Using French company data from 2018 to 2020, we found a positive relationship between AI adoption and increased hiring and sales at the company level. This finding aligns with most recent studies on automation’s effects on labor demand at the firm level and supports the view that AI adoption drives productivity improvements by helping companies expand their areas of operation.
The positive effect on productivity appears to outweigh the potential reduction in labor demand resulting from AI assuming tasks associated with specific worker types and jobs. Even in traditionally vulnerable-to-automation occupations like accounting, telemarketing, and secretarial work, AI’s impact on labor demand remains positive. While some AI applications (e.g., cybersecurity) increase employment levels, others (administrative processes) show minor negative effects. However, these differences seem to stem more from varying AI uses rather than inherent occupation characteristics.
Overall, the primary risk for workers is being replaced by those from AI-adopting companies rather than direct replacement by AI. Countries attempting to slow AI adoption may inadvertently harm local employment, as their businesses face competition from foreign counterparts that have embraced AI.
Policy Recommendations
Our data interpretation suggests that AI can drive growth and employment, but realizing this potential requires appropriate reforms. Competition policy should ensure that dominant companies in high-value AI chain segments don’t hinder new innovators’ entry. Our study also reveals that AI-adopting companies tend to be larger and more productive, suggesting dominant firms are better positioned to benefit from the AI revolution.
To prevent market concentration and entrenched market power, smaller businesses should be encouraged to adopt AI through a combination of competition policy and industry policies improving data and computing power access. Ensuring high-quality education’s wide availability, along with active labor market training programs and policies, is crucial to maximizing AI’s job creation potential and minimizing its negative effects on workers.
Conclusion
The next technological revolution is underway. The future of entire countries and economies hinges on their willingness and ability to adapt to it.
About the Authors
Philippe Aghion, Nobel laureate in Economic Sciences 2025, is a professor at the Collège de France and the London School of Economics and an associated researcher at the Centre for Economic Performance.
Simon Bunel is an economist at the Banque de France.
Xavier Jaravel is a professor of Economics at the London School of Economics.
Copyright: Project Syndicate, 1995 – 2025
www.project-syndicate.org
