Why it matters. It is one of the first central-bank studies to test whether the generative-AI boom is actually making economies more productive — a question with huge stakes for wages, investment, and growth worldwide.
Background. The Bank of Korea (BOK) is South Korea's central bank and a respected source of economic research; reports like this carry weight in domestic policy debates. South Korea is one of the world's most digitally advanced and AI-eager economies, with heavy corporate adoption of tools like ChatGPT and Gemini, making it an unusually clear test case for AI's real-world payoff.
What to watch next. Watch whether Korean firms and policymakers act on the BOK's advice to redesign workflows and incentives, which the bank says is the key to converting time savings into genuine productivity gains.
South Korea’s central bank said on June 7 that generative AI tools such as ChatGPT have trimmed about 90 minutes off the average worker’s week, yet the time saved has not translated into higher productivity — a gap the bank calls a “productivity disconnect.”
The finding comes from a report titled “Does AI Adoption Raise Productivity? An Analysis of the First Three Years,” published by the Bank of Korea (BOK), the country’s central monetary authority. It is one of the first official assessments of how generative AI is reshaping work in a major advanced economy.
Time saved, but output unchanged
According to the report, workers who used generative AI services cut their working hours by an average of 3.8% — roughly one and a half hours out of a standard 40-hour week — by deploying tools like ChatGPT and Google’s Gemini for tasks such as drafting documents and analyzing data.
The savings were uneven across occupations. Professionals, office workers, and managers saw the largest reductions, while service and manual-trade jobs benefited far less. But when the bank estimated how much overall productivity would rise if all that saved time were funneled back into output, the figure came to just about 1% — effectively negligible.
Productivity here is measured as gross domestic product (GDP) divided by hours worked. Strikingly, when researchers compared the rate of time saved against the rate of increased work output, the correlation between the two was zero. Faster work, in other words, has not meant more work getting done.
Why the gains stall
The BOK attributes the disconnect to AI’s current role as an assistant for individual tasks rather than a tool woven through the entire workflow. Speeding up data analysis does little, the report notes, if the surrounding steps — organizational decision-making, approvals, and sign-offs — remain as slow as before.
The bank also warned of a human factor: if firms do not reward the efficiency gains appropriately, AI adoption could actually reduce workers’ effort and hold productivity back further.
Importantly, the BOK framed this not as a limitation of the technology itself but as a typical transition phase seen in the early years of any general-purpose technology. The decisive factor, it argued, is how companies use AI and restructure their organizations around it.
What the central bank recommends
The report urges firms to draw a clear line between standardized tasks and open-ended tasks. Routine work — summarizing reports, tidying data — should be handed to AI, while open-ended work requiring human experience, judgment, and creativity should be done in collaboration with AI.
“What matters going forward is not AI adoption itself, but continuously monitoring how AI actually connects to changes in productivity,” the researchers wrote. They added that policy should focus on systematically supporting the organizational and institutional shifts needed to turn AI use into real productivity growth.
