AI Is Said to Be Transforming the World... Has Productivity Really Increased? [Weekend Money]

Fed Chair Also Asked the 'AI-Productivity' Question Last Month
NH Investment & Securities: "There Is a Time Lag Between New Technology and Productivity"
"Inflection Point for Acceleration and Visibility Will Be Around 2028"
Similar Time Lag Observed in Electrical and Digital Revolutions

Nearly four years have passed since the launch of ChatGPT 3.5, and artificial intelligence (AI) is now widely used as a common tool not just by individuals but also at the organizational level. Companies are eager to tout their rates of AI adoption and usage; however, when looking at key indicators such as economic growth and productivity, there appears to be little actual change.

Even Kevin Warsh, Chair of the Federal Reserve (Fed), raised this question. In his speech at Jackson Hole on August 28, he asked, "Will AI bring meaningful and sustained productivity gains across the economy, and if so, when will that happen?"

NH Investment & Securities attempted to answer this question in a recently published report.

NH Investment & Securities researcher Kim Yong stated, "For AI technology to translate into accelerated productivity, there must first be intangible investment, such as intellectual property, organizational capital expenditure, and employee redeployment." He explained that, considering the expanded spending in these areas since mid-last year, there is a high probability that the inflection point for productivity acceleration will be around 2028.

Even Good Technology Requires Changes to the 'System' to Be Effective

There is inevitably a time lag between technological advancements from AI and actual improvements in productivity. Imagine a factory introducing new machinery. No matter how advanced the machinery is, unless workflows are redesigned and employees are retrained to suit the equipment, efficiency will not improve. In fact, in the beginning, the adjustment period could lead to greater confusion and even slower operations.

Kim refers to this overall adaptation process as "intangible investment." This concept includes costs associated with consulting for organizational restructuring, investment in intellectual property such as software and research & development (R&D), and reassigning employees to new roles. No matter how rapidly technology advances, these intangible investments must be made first for productivity to ultimately increase.

History supports this view. During the internet revolution of the 1990s, computer performance improved most rapidly in the early 1990s. But it was not until the late 1990s that U.S. productivity made a significant leap. During those intervening years, companies poured money into software and engaged management consultants to overhaul their operations for the internet era. The same was true over a century ago with the introduction of electricity. Factories spent more time reconfiguring equipment layouts to utilize electric power than adopting the technology itself, and only after these adjustments were productivity gains realized.

A similar pattern was repeated during the COVID-19 pandemic. The shift to remote work and large-scale redeployment of workers between 2020 and 2021 led to a significant surge in U.S. labor productivity starting in 2023 and continuing into 2024. The International Monetary Fund (IMF) estimated that 80% of this productivity growth during the period was attributable to employee redeployment. The time lag was roughly two years.

AI Is in the 'Adaptation Phase' Now... 2028 Will Be the Watershed

It is anticipated that this pattern will also appear in the current era of AI.

The pace of technological advancement in AI itself is unprecedented. AI evaluation agency METR uses the "time horizon," or the maximum length of task AI can accomplish independently, as a performance metric. This metric increased tenfold in just one year. For a human analogy, it is as if a new employee who could barely handle a one-hour task alone has advanced in a year to managing a 10-hour project independently.

In contrast, intangible investment to reorganize businesses for AI only started to increase from mid-2025. The hiring of management consultants and job postings related to AI rose noticeably beginning in 2026, and software and R&D investments also started to climb again from the middle of last year. The number of new business startups in the U.S. has also surged during this period, suggesting an AI-related impact.

Given the historic pattern that productivity gains occur about two years after peak intangible investment, Kim calculates that visible acceleration in productivity is likely to become apparent around 2028. This means it is premature to dismiss the AI revolution as overhyped merely because current economic indicators related to AI seem lackluster.

There are also inherent measurement limitations. Current statistical methods do not accurately reflect improvements in AI-powered service quality in inflation or growth rates. For example, there are statistics showing that software prices rose by about 20% year-on-year, but this does not capture the corresponding "value increase" due to AI enhancements. If such improvements were included, the actual inflation rate would likely be lower and consumption growth higher than currently suggested figures.

Kim pointed out, "As the AI sector grows larger, issues with measurement errors such as quality adjustments in services become more significant," but added, "Actual productivity growth may be occurring more rapidly than early measurements indicate."

This content was produced with the assistance of AI translation services.