Trends in employment and artificial intelligence in the labor market

The full study is currently available only in Hebrew.

Artificial intelligence is developing and being implemented at a record pace. Not only does it generate substantial public interest but it also gives rise to fundamental questions and serious concerns about the future of the labor market. A new Taub Center study analyzes the impact of these new technologies on workers in Israel, based on labor force survey data from 2023–2024.

The characteristics of exposure to artificial intelligence, as identified in the study, enable an assessment and projection of which occupations in Israel will be more and which will be less affected by AI. Hence, it is possible to forecast which populations are likely to “benefit” from it and which may be harmed.

The study found that between 2023 and 2024 there was an increase of at least five percentage points in the average worker’s level of exposure – a growth explained primarily by technological advancement rather than by changes in the labor market. The sharpest increase in exposure was recorded among occupations characterized by a high risk of replacement by artificial intelligence.

The researchers, Michael Debowy, Jonathan Winter, Prof. Gil Epstein, Prof. Avi Weiss, and Efrat Behar‑Netanel, call for the formulation of policy tailored to cope with these changes.

Exposure levels
In order to estimate the share of individuals exposed to artificial intelligence in a given industry or population, several measures have been developed. The exposure score is calculated based on the scope of tasks that artificial intelligence purports to perform out of all the tasks comprising the occupation. The score assigned to each occupation reflects the percentage of that occupation’s typical tasks that artificial intelligence can carry out with human‑level quality, at a speed exceeding that of a human worker.

Replacement or complementarity? It all depends on the occupation
The study sharpens the distinction between occupations in which artificial intelligence complements human workers and assists them, and those in which it may replace them. The sharpest increase in exposure levels between 2023 and 2024 was recorded in occupations with “low complementarity” — namely those in which human intervention is not necessary to complete the task, and, therefore, technology may replace the worker. The financial, administrative, and commerce sectors are notable for high exposure levels and particularly high substitutability. In contrast, occupations such as teaching and medicine, characterized by “high complementarity,” are also exposed to the technology, but workers in these fields are expected to be assisted rather than replaced by it.

Women show higher exposure levels to AI than men, and particularly in those occupations with a higher risk of replacement. The gaps are particularly notable in the Arab sector
The study found that the exposure rate of female workers is higher than that of male workers. However, female workers are “polarized” on the complementarity index, with a relatively high share concentrated at the extremes of the distribution. The main source of the exposure gap between the genders arises from occupations at the highest risk of replacement: more than one‑quarter of women fall into this group compared with less than one‑fifth of men, meaning that women’s exposure rates are significantly higher than those of their male counterparts (in part due to women’s concentration in occupations such as secretarial roles, sales, and telephone customer service). This pattern holds within each sector as well, and the largest gender gap is in the Arab sector, where female workers’ exposure rate is twice that of male workers on average — although exposure rates among Arab women and Arab men remain significantly lower than for Jewish women and Jewish men.

High exposure rates to AI raise the risk of remaining outside the labor market
The study shows that average exposure rates differ between employed and unemployed groups, and that among self‑employed individuals and those not participating in the labor market, the exposure rate is somewhat lower than among employees and the unemployed. Moreover, the AI exposure index predicts non‑employment: that is, an individual’s probability of being unemployed or discouraged from seeking employment rises significantly with increased exposure to artificial intelligence. This finding may indicate the onset of a tangible impact of artificial intelligence on employment levels in Israel.