Companies that use AI as a collaborative tool are experiencing steady hiring, while those that fully automate tasks are seeing a decline in entry-level job openings, according to new data. This marks a significant shift from previous discussions that primarily focused on theoretical frameworks regarding automation and augmentation.
On August 26, the Stanford Institute for Digital Economy, in collaboration with ADP Research, released the 'Canary Dashboard.' The report indicates that employment among entry-level workers aged 22 to 25 in high AI exposure roles has decreased by 3.8% annually, while those in low AI exposure roles have seen a 2% increase.
This gap has widened from 2.8% in April 2024 to approximately 4% currently, with a consistent monthly increase of about 0.5 percentage points over nearly four years. Professor Erik Brynjolfsson noted, "Whatever the cause, this trend shows no signs of abating."
One of the most striking findings from the study is the cross-analysis with Anthropic's AI usage data. Researchers found that the decline in entry-level employment is only evident in sectors that utilize AI for automation, while those that employ AI for augmentation do not experience similar reductions.
In other words, organizations that use AI as a supportive and collaborative tool are not significantly affected in their hiring practices, whereas those that delegate entire tasks to AI are reducing their entry-level hiring. The researchers emphasized that how AI is utilized is crucial.
This empirical evidence aligns with academic frameworks that distinguish between automation and augmentation. The Stanford research team, which includes Professor Brynjolfsson, previously presented a framework assessing the automation and augmentation potential of AI across 844 tasks and 104 occupations using the U.S. Department of Labor's O*NET database.
Jobs with high automation potential face significant replacement risks with AI implementation, while those with high augmentation potential allow AI to handle simple, repetitive tasks, leaving room for workers to create added value.
In this context, South Korea is rapidly transitioning from automation to augmentation in its AI usage. An analysis released earlier this year by the Korea Institute for International Economic Policy (KIEP) using Anthropic's economic indicators revealed that the proportion of AI automation in South Korea dropped from 44.5% in August of last year to 38.8% this year, a decline of 5.7 percentage points. Conversely, the augmentation rate increased from 55.5% to 61.2%.
During the same period, Japan (43.3% to 38.6%), Taiwan (46.6% to 40.9%), and Singapore (45.3% to 39.6%) also saw declines in automation rates, with South Korea experiencing the largest drop among the four countries.
International comparisons consistently reflect this trend. According to an economic indicators report released by Anthropic last year, countries with lower income levels and AI adoption rates tend to prefer full task delegation through automation, while those with higher income levels and adoption rates are increasingly favoring augmentation methods focused on learning and repetition. Even after controlling for differences in job composition across countries, a correlation was found indicating that for every 1% increase in AI usage per capita, the proportion of automation decreases by about 3%.
* This article has been translated by AI.
Copyright ⓒ Aju Press All rights reserved.