Employees using AI are working faster, but the economy isn't more efficient. A look at what happened in the pre-Internet era might explain why

In 2026, two intriguing phenomena are emerging within the economy. On the one hand, economic expansion continues robustly even as job growth decelerates, suggesting that productivity among current employees is on the rise. Conversely, various metrics indicate that productivity growth has barely progressed in recent years, showing a marked slowdown in the first quarter of 2026. These situations typically seem mutually exclusive.

Proponents of technology posit that AI has the potential to optimize workflows and significantly enhance the productivity of the U.S. economy—a metric reflecting the efficiency with which labor and other resources convert into goods and services. Although this growth has yet to be evident in official data, AI may explain the current inconsistencies in productivity statistics.

Research conducted by the London School of Economics last year indicates that employees utilizing AI are more likely to accomplish the same quantity of work in less time, effectively saving an entire workday per week. Economists refer to this phenomenon as capital deepening, where access to superior tools leads to increased individual productivity, much like a construction worker upgrading from a shovel to a mechanical excavator.

An increased understanding of this process may be drawn from a recent research brief published by the Federal Reserve Bank of San Francisco. Similar to the significant investments in AI integration seen today, economists examining the nascent days of the Internet in the early to mid-1990s were also left perplexed. Employees had newly acquired access to transformative technology, yet many businesses remained entrenched in a "productivity paradox" that impacted the U.S. from the 1970s through the 1990s, during which massive investments in IT didn't translate into improved efficiency.

This stagnation ultimately served as a temporary delay, and should history repeat itself, the U.S. economy may be experiencing the initial phases of a substantial productivity surge without immediate acknowledgment.

"Determining whether a prolonged period of high growth has begun is challenging in real time and is often only identifiable with hindsight," the Federal Reserve researchers noted.

Fickle productivity

Economists typically employ two primary metrics to assess productivity, which are currently indicating conflicting trends. The first, labor productivity, examines output per labor unit. The second, total factor productivity (TFP), offers a broader perspective, measuring how effectively the entire economy converts inputs into outputs.

In recent years, labor productivity has shown solid improvements, whereas TFP has encountered challenges in achieving significant growth following a post-pandemic spike. The Federal Reserve researchers interpreted this discrepancy as evidence that employees are working more rapidly and productively at an individual level; however, the overall workforce has not necessarily become more efficient.

This occurrence aligns with developments seen during the computer and Internet boom of the 1990s. Beginning in mid-1996, labor productivity started to accelerate at a rate exceeding that of TFP, but the comprehensive productivity benefits associated with the Internet didn't fully materialize in aggregate data until several years later.

Nobel laureate Robert Solow captured this disconnect succinctly in a statement that has become well-known: "You can see the computer age everywhere but in the productivity statistics," he remarked in 1987.

A similar scenario is unfolding today, with experts such as Torsten Slok, chief economist at Apollo, applying Solow's insights to the era of AI. Business investments in AI are surging as companies anticipate a forthcoming productivity boom, suggesting that each worker is gaining access to an expanding array of tools that have yet to be efficiently integrated throughout the economy.

The challenges accompanying AI adoption are brought to light by various studies. A Harvard Business Review analysis of 200 employees at a U.S. technology firm revealed that while those using AI tools did save time on their tasks, this time was often reallocated to other responsibilities, resulting in fewer breaks overall. Consequently, most workers ended up working longer hours, increasing the risk of burnout. Another study from Harvard highlighted that extensive AI utilization at work might lead to excessive cognitive demands, contributing to instances of "brain fry."

Furthermore, a study conducted by the Atlanta Fed in March provided more pointed insights. The survey, encompassing approximately 750 corporate executives, generally indicated that productivity is indeed improving due to AI. However, the perceived productivity improvements reported by executives surpassed measurable indicators such as company revenue, which the Fed attributed to "delayed output realizations."

While workers may perceive themselves as more productive with AI—often rightly so—the lack of a discernible impact on the economy as a whole reflects notable parallels to the early Internet era, wherein data initially failed to forecast an impending productivity boom.

"If contemporary developments resemble those of the mid-1990s, we could be at the infancy of a productivity explosion fueled by AI that will become evident only in retrospect," the San Francisco Fed researchers stated.