AI Is Saving Korean Workers Time. It Isn’t Making Them More Productive — Yet
GGMA | MARKETS & ECONOMICS
Bank of Korea finds a ‘productivity disconnect’: generative AI trims roughly 1.5 hours a week off the average workweek, but the saved time shows almost no link to higher output.
By GGMA Staff
SEOUL — Generative artificial intelligence is giving Korean workers back time, but so far that time is vanishing rather than turning into measurable output, according to new research from the Bank of Korea.
AI adoption has cut average work time by about 3.8%, or roughly 1.5 hours a week, the central bank said in an Issue Note published June 9. Yet those hours show essentially no correlation with faster output growth — a gap the authors call a “productivity disconnect” that helps explain why the long-promised AI productivity boom has yet to surface in the macro data.
The finding lands at the center of one of the biggest open questions in the global economy: whether the rapid spread of tools like large language models will lift growth, or merely reshuffle how the workday is spent. Donghyun Suh, Samil Oh and Jongwon Yoon, the economists behind the study, drew on a household survey to trace AI’s effects through the first three years of mass adoption.
Their conclusion is that AI is stuck in what they describe as an “efficiency” stage — faster individual tasks — and has not yet crossed into a “productivity” stage, where those gains feed through to greater output.
The math that doesn’t add up
The potential is real on paper. If every hour AI frees up were redeployed into productive work, the time savings would translate into a productivity gain of about 1.0%, the authors estimated. The largest time savings accrued to lower-skilled workers and to the heaviest AI users.
The problem is conversion. The hours saved at the desk are not reappearing as additional output, the study found. AI has sped up discrete tasks — drafting, summarizing, coding — without yet reshaping the workflows, organizational structures and labor allocation that determine whether faster tasks become a more productive firm.
In other words, the technology has optimized the step but not the system.
Who is bucking the trend
Not everyone is leaving the gains on the table. The self-employed, professionals and the most intensive AI users did register genuine productivity improvements, the authors found. What those groups share is strong performance incentives and a high degree of control over how they work — pointing to organizational design and incentive structures, rather than the technology itself, as the binding constraint.
That distinction matters for policy and for employers. It suggests the payoff from AI accrues to those positioned to redesign their own work around it, and stalls where rigid processes and weak incentives leave saved time unclaimed.
A familiar pattern
The Bank of Korea frames the disconnect not as a failure but as a stage. Early in the life of a general-purpose technology, gains routinely lag adoption — the so-called J-curve, where output dips or flatlines before it climbs, and the Solow Paradox, the observation that the computer age was visible everywhere except in the productivity statistics.
By that reading, the question is not whether AI raises productivity, but when — and that timing will hinge on choices made by companies and policymakers rather than on the pace of the technology.
To convert efficiency into productivity, the authors argued for redesigning work processes and organizational structures, distinguishing standardized tasks from open-ended ones, reallocating labor toward higher-value work, and building performance-based incentive systems. They also urged continued monitoring of how AI is reshaping the skill-formation pathways of younger workers — a shift with consequences for the labor force that long outlast the current cycle.
For now, the headline is a paradox in plain numbers: the tools are working, the time is being saved, and the productivity gains have yet to show up where they count.
Source: Bank of Korea, “Does AI Adoption Improve Productivity? Effects Over the First Three Years,” BOK Issue Note No. 2026-12, June 9, 2026, by Donghyun Suh, Samil Oh and Jongwon Yoon.