AJP Asia Intelligence Agency

Two-Thirds of South Korean Workers Use AI at Work, Saving 3.7 Hours Weekly

Two out of three South Korean workers are utilizing generative artificial intelligence (AI) in their jobs. The amount of work time saved through AI has more than doubled in the past year. However, the reduction in working hours has not yet translated into significant productivity improvements.


According to a report titled 'AI Diffusion and Productivity Effects: Wider, More Efficient, but Still Shallow' released by the Bank of Korea on November 17, the rate of generative AI usage among domestic workers rose from 51.8% last year to 66.5% this year, an increase of 14.7 percentage points.


This rate is nearly double the average of 33.7% among six major countries in the U.S. and Europe, making it the highest among the countries compared. The AI usage rate in the U.S. stands at 43.0%, which is 23.5 percentage points lower than that of South Korea.


The proportion of workers who regularly use AI also increased from 17.1% last year to 30.6% this year. The average weekly usage time for workers who utilized AI in their jobs rose from 5.9 hours to 8.0 hours, a nearly 36% increase.


The time-saving effect of using AI has also significantly expanded. The average reduction in working hours for employees using AI increased from 3.8% last year to 9.2% this year, a 2.4-fold increase. Based on a standard 40-hour workweek, the time saved has grown from 1.5 hours to 3.7 hours.


Notably, workers who use paid services experience greater time-saving effects. The average reduction in working hours for free users was 8.3%, while individual paid users reported a reduction of 11.4%, and those whose companies subsidize the costs reported 11.6%.


Among workers using AI, 28.4% subscribe to paid services. The average monthly expenditure is 49,000 won, which estimates the annual AI usage fees for all domestic workers at approximately 2.8 trillion won.


The Bank of Korea's research team estimates the potential value of the time saved through AI usage at 61.1 trillion won annually. Assuming all saved time is reinvested into productive activities, the potential output increase effect is estimated at 2.6%, surpassing last year's estimate of 1.0%. However, this does not imply actual income growth or GDP growth effects.


The challenge remains that the reduction in working hours through AI has not sufficiently connected to actual productivity improvements.


Responses indicating that time saved through AI is focused on core tasks accounted for 33.8%, while those using it for additional tasks accounted for 29.3%. Combined, these responses reached 63.1%, but the correlation coefficient between the reduction in working hours and the increase in work output was only 0.08. The correlation coefficient between the reduction in working hours and quality improvement was also low at 0.09.


Despite the widespread adoption of AI, its application to core job functions remains inadequate. The analysis suggests that the time-saving effects achieved through actual AI usage fall significantly short of what is technically possible.


However, workers with a higher proportion of core tasks tend to show relatively greater improvements in work quality. This indicates that the benefits of AI usage may manifest more in enhancing the quality of outputs rather than in the speed or quantity of work processed.


Oh Sam-il, head of the employment research team at the Bank of Korea, stated, "Overall, AI is widely used for non-core tasks, but those who achieve quality improvements are ultimately those who use it for core tasks. The quality improvement effects are seen among those who pay for their own usage and those who apply it to core tasks."


There are also opinions that for AI usage to lead to productivity improvements, companies must also change their work methods and organizational operations. Oh emphasized, "For technology to realize its potential, there must be complementary assets such as organizational structure, workflow, and compensation systems that can support the technology. When these complementary assets are in place, it will be possible to utilize AI closer to its theoretical potential."





* This article has been translated by AI.