New research from the Federal Reserve Bank of St. Louis indicates that while optimism surrounding AI’s impact on productivity is prevalent among corporate executives, measurable gains remain elusive, aligning with historical patterns observed in technological advancements.
The Federal Reserve Bank of St. Louis has released a significant report analyzing nearly 490,000 corporate earnings calls, revealing that artificial intelligence (AI) has yet to produce a measurable increase in aggregate productivity. This finding aligns with previous data trends observed over the past three years, suggesting that while executives express optimism about AI’s potential, tangible results have not yet materialized. The research presents a thought-provoking theory: AI may be creating structural changes that obscure real productivity gains by simultaneously reducing the value of outputs.
Understanding the Productivity Paradox
One of the authors of the study, Serdar Ozkan, highlighted that as AI technology lowers the cost of producing certain outputs, the overall value of those outputs also diminishes. This phenomenon suggests that while production becomes easier and cheaper, the gains may not be reflected in productivity statistics. “Some things are going to become more abundant,” Ozkan stated, “That means they’re also going to become probably less valuable.” This perspective underscores a critical aspect of economic productivity: the interplay between production costs and market value.
In-Depth Data Analysis
The research team, including Ozkan and co-author Aakash Kalyani, examined earnings call transcripts from 5,198 publicly traded firms in the United States between 2000 and 2025. They utilized AI models to categorize discussions regarding productivity and AI. Notably, the proportion of commentary related to AI has surged from virtually zero prior to the introduction of ChatGPT in late 2022 to approximately 15% of all productivity discussions by the end of 2025. Furthermore, about 95% of these AI-related comments focus on anticipated future gains, as opposed to gains that have already been realized, a trend that has remained consistent since 2023. In contrast, when executives discuss productivity in general terms, 75% express a positive outlook.
Historical Context of Technological Advancements
Ozkan referenced economist Robert Solow’s famous observation that while the computer age was visible, its benefits were not reflected in productivity statistics. This historical analogy draws parallels to the current state of AI, where the realization of productivity gains may require time for firms to adapt, retrain employees, and restructure workflows. He noted that it took decades after the introduction of electrification for its benefits to become apparent in productivity figures.
Stanford economist Erik Brynjolfsson has also contributed to this discourse, describing the present scenario as the modern sequel to the “productivity paradox” he identified in a 1993 paper. Kalyani further emphasized that while excitement about AI is palpable, the consensus among economists is that substantial gains will manifest in the future, rather than immediately. He explained that the diffusion of technology across various sectors typically unfolds over several decades, indicating that the rapid adoption of AI might not deviate from this trend.
Current Productivity Trends and AI Investment
Supporting the findings of the St. Louis Fed, a separate analysis from the Kansas City Fed has concluded that recent productivity increases are not widespread, with a limited number of industries capturing most of the gains. Additionally, Fed Chair Kevin Warsh addressed Congress in July, noting that AI had not yet displaced workers but had slightly enhanced their productivity. However, he cautioned that the long-term impacts of AI on the workforce may take considerable time to materialize.
Previous research from the St. Louis Fed estimated that generative AI contributed only a 1.1% increase in productivity by late 2024, relative to the figures from 2022. This increase is modest compared to the overall productivity growth rates of 2.3% and 1.6% recorded in 2024 and 2023, respectively.
Investment Trends Reflecting Optimism
Despite the lack of immediate results, the researchers emphasized that corporate enthusiasm for AI is translating into actual investment. Kalyani pointed out that firms expressing positive sentiments about AI have significantly increased their research and development (R&D) expenditures and capital investments. This trend marks a shift from earlier studies that indicated minimal correlation between positive AI discussions and investment patterns. The San Francisco Fed also observed that firms optimistic about AI saw greater R&D growth by 2025 compared to their counterparts.
The Limitations of Productivity Metrics
Ozkan’s argument regarding the abundance created by AI highlights a crucial limitation in measuring productivity. While AI enhances certain functions, such as research and drafting, it does not eliminate bottlenecks that persist in organizational processes. For instance, basic logistical tasks, such as scheduling meetings, continue to operate at the same pace as they did four years ago. Thus, productivity cannot be distilled into a single figure; it represents a complex output resulting from various interconnected processes, of which AI has only accelerated certain segments.
When asked what metrics might eventually confirm AI’s contributions to productivity, Kalyani acknowledged the uncertainty inherent in such predictions. The advent of impactful applications often emerges unpredictably, evolving through a process that may seem chaotic yet ultimately aggregates into significant advancements.
This nuanced understanding of AI’s potential impact on productivity reveals that while corporate executives may rightfully anticipate future benefits, the actual manifestation of these gains—as reflected in productivity statistics—may remain elusive until they have already occurred.
