OIT - Generative AI and jobs - A 2025 update

 ILO Brief 1  Research Brief May 2025

Generative AI and jobs: A 2025 update Paweł Gmyrek (ILO1), Janine Berg (ILO1), Karol Kamiński (NASK-PIB), Filip Konopczyński (NASK-PIB2), Agnieszka Ładna (NASK-PIB), Balint Nafradi (ILO), Konrad Rosłaniec (NASK-PIB), Marek Troszyński (NASK-PIB, Civitas University)

 Updates ILO’s 2023 estimates of potential occupational exposure to generative AI ( GenAI) technology and the employment shares of affected occupations.  Incorporates a more refined m ethodology that draws on b oth human and AI insight , and which is assessed at the 6-digit occupational level covering nearly 30,000 tasks.  Defines four progressively increasing gradients of GenAI exposure depending on the mean exposure score and the degree of task variability for each ISCO-08 occupation.  Overall, the automation scores are slightly lower than in 2023 (a mean automation score of 0.29 in 2025 versus 0.30 in 2023), though the variability of scores is considerably lower (standard deviation 0.14 in 2025 v. 0.30 in 2023).

 Growing abilities of GenAI models in such areas as voice, image and video generation have increased automation scores for a range of tasks in mediaand web-related occupations.  One in four workers across the world are in an occupation with some degree of GenAI exposure, but because of the continued need for human input, most jobs will be transformed rather than made redundant.  There is a need to ensure that the transition is managed through social dialogue , to enhance both working conditions and productivity.

but because of the continued need for human input, most jobs will be transformed rather than made redundant.  There is a need to ensure that the transition is managed through social dialogue , to enhance both working conditions and productivity.

1 Department of Research, ILO Geneva. 2 NASK-PIB is a National Research Institute under the supervision of the Ministry of Digital Affairs in Poland. 3 Department of Statistics, ILO Geneva. Key points ILO Brief 2  Introduction In August 2023, the ILO published the first global employment estimates of potential occupational exposure to generative artificial intelligence (GenAI) (Gmyrek et al., 2023). Since then, there have been further advancements in the technology , but also in workers’ and employers’ understanding of both the potential and the limitations in the use of this technology in currently existing jobs. This policy brief summarizes the findings from Working Paper 140 (May, 2025) which improves upon the 2023 methodology to update occupational exposure scores and associated global employment estimates. The study was conducted by the ILO in partnership with NASK-PIB, the National Research Institute under the Ministry of Digital Affairs in Poland, to develop a more refined assessment of potential effects of GenAI on employment across countries. Central to this work, and the earlier 2023 work, is the insight that jobs are a “bundle of tasks” ( Autor, 2015). As such, task automation might, or might not lead to job automation, depending on the importance of a particular task to an occupation, as well as the degree of task variability within that occupation. Moreover, the estimates reflect “potential occupational exposure” to the technology – essentially a theoretical assessment of what tasks can currently be performed using the technology – rather than its application in practice, which may be constrained by inadequate infrastructure or skills, high technological costs, or competing organizational priorities. As in 2023, the objective of the exercise is not to have precise figures, but rather to provide insights into possible directions

technology – rather than its application in practice, which may be constrained by inadequate infrastructure or skills, high technological costs, or competing organizational priorities. As in 2023, the objective of the exercise is not to have precise figures, but rather to provide insights into possible directions of change. Our aim is to encourage governments and social partners to proactively design policies that support orderly, fair, and consultative transitions, rather than dealing with change in a reactive manner.

 Methodology This study improves the methodology developed in 2023, by incorporating additional sources of information and introducing an extra layer of expert verification. The 2023 methodology used task descriptions from the 4 -digit International Standard Classification of Occupations (ISCO-08) and the GPT -4 model to estimate task -level and occupational scores of exposure to GenAI. Figure 1. Schema of the scoring stages ILO Brief 3 This study builds on the 2023 work, but it begins with a detailed national occupational classification. By using Poland’s 6digit system, we increase the number of tasks analysed to over 30,000 – ten times the amount provided at the 4-digit ISCO08 level. In addition, the methodology incorporates a human-based evaluation, by surveying 1, people currently employed in each ISCO-08 1-digit group and asking them to rank the potential for automating tasks with GenAI technology, based on a representative sample of tasks that pertain to their occupational group (step 2, Figure 1). This is followed by a detailed review of a large sample of tasks assessed in the survey, conducted by national and international experts who validate or adjust the task automation scores attributed by the survey participants (step 5). We then input these adjusted scores into an AI model that generates new scores incorporating this human judgement, covering nearly 30,000 individual tasks

pertaining to 2,500 occupations at the 6-digit level in Poland (step 9). Finally, we re-generate scores for the tasks in ISCO08, previously provided by Gmyrek et al. (2023), thereby adjusting the 2023 ILO GenAI Exposure Index.2

 Occupational Exposure To classify the exposure of ISCO-08 occupations to GenAI, we update the framework introduced in 2023. As in the original approach, we rely on two moments of distribution : occupational mean and standard deviation (SD) of task -level scores pertaining to a given 4 -digit occupation. However, instead of using automation and augmentation potential as two extreme categories with the "big unknown" in between, we now adopt a more nuanced classification scheme that represents a spectrum of exposure, divide into four progressively increasing gradients. Gradient 1 represents occupations with low overall GenAI exposure but significant variability across tasks (Low exposure, high task variability). While some tasks within these roles may have high automation potential, the occupation has many tasks that continue requiring human roles , aligning closely with the notion of augmentation in the old framework. Gradient 2 includes occupations with moderate GenAI exposure and a mix of highly exposed and minimally exposed tasks, resulting in uneven impacts where some tasks may be disrupted while others remain unaffected (Moderate exposure, high task variability) . Gradient 3 captures occupations where a significant portion of tasks are consistently exposed to GenAI, signalling growing automation risks and requiring adaptation strategies for workers (Significant exposure, high task variability) . Gradient 4 highlights occupations with the highest share of tasks exposed to potential GenAI-driven automation, and with a high consistency of this exposure across tasks (Highest exposure, low task variability) . This gradient aligns closely with the notion of automation risk in our 2023 classification. In addition, among o ccupations outside these four categories, we introduce a more granular distinction of Minimal Exposure, where some interaction with GenAI may occur , but without significantly altering the nature of these roles , and the lowest category of Not Exposed occupations, where GenAI currently has no observable potential to automate tasks. (See Figure 2)

outside these four categories, we introduce a more granular distinction of Minimal Exposure, where some interaction with GenAI may occur , but without significantly altering the nature of these roles , and the lowest category of Not Exposed occupations, where GenAI currently has no observable potential to automate tasks. (See Figure 2) This revised framework addresses some limitations of the earlier categorization system, particularly the interpretative challenges posed by the category “the big unknown". By expanding the exposure categories into progressively increasing gradients, the updated framework improves the readability and interpretability of results . It also captures, in a more nuanced way, how GenAI can impact occupations at varying levels of exposure, based on task-level variability and overall occupational risks. Furthermore, the framework's flexibility enhances its applicability to country-specific contexts, allowing policymakers to better understand the distribution of GenAI's potential impact and prioritize interventions for the most affected groups in the national contexts . We stress that these classifications are only illustrative, since any type of task scoring system carries a degree of subjectivity, and since the abilities of GenAI and derivative technologies evolve rapidly.

2 For more details on the methodology, see ILO Working Paper no. 140 (https://www.ilo.org/publications/generative-ai-and-jobs-refinedglobal-index-occupational-exposure). ILO Brief 4 Figure 2. AI Exposure Gradients: ISCO-08 4-digit level occupations

Table 1: GenAI Exposure Gradients by Occupation GenAI Exposure Definition (Mean & SD of Task Scores) Interpretation

Exposed: Gradient 4

(Highest exposure, low task variability) μ ≥ 0.6 and μ - σ >= 0.5 High and consistent GenAI exposure across tasks within the occupation. Most current tasks in these jobs have a high potential of automation, with little variability in task-level exposure.

Exposed: Gradient 3

(Significant exposure, high task variability) 0.5 ≤ μ < 0.6 and μ + σ ≥ 0.5 Above-moderate occupational exposure: even though some tasks

with little variability in task-level exposure.

Exposed: Gradient 3

(Significant exposure, high task variability) 0.5 ≤ μ < 0.6 and μ + σ ≥ 0.5 Above-moderate occupational exposure: even though some tasks remain less exposed, the overall potential of automation of the current tasks with GenAI is growing in these occupations.

Exposed: Gradient 2

(Moderate exposure, high task variability) 0.4 ≤ μ < 0.5 and μ + σ ≥ 0.5 Moderate occupational AI exposure, with high task-level variability. These occupations include a mix of some tasks that are exposed to GenAI and others not at risk, making the impact uneven.

Exposed: Gradient 1

(Low exposure, high task variability) μ < 0.4 and μ + σ ≥ 0.5 Low overall GenAI exposure at the occupational level, but high variability across tasks. Some tasks within these occupations have an elevated automation potential, even if the occupation as a whole remains strongly reliant on tasks that have a low potential of automation. Low Exposure (Minimal exposure, moderate task variability) μ < 0.5 and μ + σ > 0.4 Occupations with low GenAI exposure, where some tasks show moderate automation potential, but overall occupational exposure remains limited. Not Exposed Occupations that don’t meet any of the above conditions. Occupations where most tasks remain relatively unaffected by GenAI, with low task variability and a stable low mean exposure score. ILO Brief 5  Comparison with 2023 exposure scores The 2025 estimates are broadly consistent with those of the ILO 2023 study, although the overall automation scores are slightly lower, with a mean automation score of 0.29 for all occupations in 2025 compared to 0.30 in 2023. However, the score dispersion within occupations is considerably lower, resulting in a more concentrated overall distribution (Figure 3).

slightly lower, with a mean automation score of 0.29 for all occupations in 2025 compared to 0.30 in 2023. However, the score dispersion within occupations is considerably lower, resulting in a more concentrated overall distribution (Figure 3). Figure 3. AI Exposure: Comparison of 2025 with 2023, ISCO -08 4-digit level occupations

As in the 2023 estimates, clerical occupations continue to exhibit the highest levels of exposure. However, there are some notable changes in the distribution of occupational means (Figure 4). Several occupations that previously received the highest scores have seen a decline in their mean scores , which highlights the distinction between the theoretical automation potential of a task and the practical insights gained from two years of experimentation with GenAI tools. While tasks such as taking meeting notes or scheduling appointments can significantly benefit from GenAI support, our 2023 scores, which for some tasks reached as high as 0.9, reflected an overly optimistic assessment of full automation potential. For the 2025 estimates, the highest task level score is 0.76, and the highest occupational mean is 0.7 (Gradient 4), meaning that there are still some tasks even within these higher-risk occupations that require human input.3

3 For a full list of the scores for each 4-digit ISCO-08 occupational group, see the Appendix of Working Paper no. 140 (https://www.ilo.org/publications/generative -ai-and-jobs-refined-global-index-occupational-exposure). ILO Brief 6 Figure 4. Difference in means for ISCO-08 4-digit level occupations with highest changes between 2023 and 2025

Nevertheless, several highly digitized occupations – such as web and media developers, statistical and database specialists, and financial and software -related roles – saw an increase in their mean scores when compared with 2023 . This rise is partly attributable to the rapid expansion of AI capabilities since our previous study. At that time, large language

Nevertheless, several highly digitized occupations – such as web and media developers, statistical and database specialists, and financial and software -related roles – saw an increase in their mean scores when compared with 2023 . This rise is partly attributable to the rapid expansion of AI capabilities since our previous study. At that time, large language models (LLMs) were primarily viewed as advanced text generators. Since then, they have acquired multimodal abilities, allowing them to process and generate text, images, audio, and video. Additionally, these models have been gaining some agentic capabilities, allowing them to execute multi -step tasks autonomously, interact with software environments, and make decisions based on contextual inputs. These advancements significantly broaden the scope of tasks that can be at least partially automated using GenAI, with software development and multimedia sectors among the leading adopters. It is important to keep in mind, nonetheless, that as in past technological transitions, new roles within occupations and entirely new occupations are likely to emerge alongside technological advancements. Key questions – largely dependent on the policies shaping this evolution – are to what extent these jobs can retain and retrain existing workers; and how will the transformation in roles within an occupation affect job quality. Will workers have more time to do creative work or will the automation lead to greater standardization of their roles and reduced autonomy? ILO Brief 7  Exposed occupations as a share of employment and comparison with 2023 estimates The next stage of our analysis takes the 112 occupations identified as exposed (Gradients 1-4) and estimates the share of employment that these occupations represent at the global, regional, and income -based levels. We apply the same method as in Gmyrek et al. (2023) , which relies on the ILO’s global estimation model. This model is based on ILO’s harmonized microdata collection and a hierarchical imputation process, integrating data throughout the ISCO -08 structure from the most reliable sources at the highest level of detail available.4 At the global level, about a quarter of all employment falls into one of the four exposure gradients, with significant

harmonized microdata collection and a hierarchical imputation process, integrating data throughout the ISCO -08 structure from the most reliable sources at the highest level of detail available.4 At the global level, about a quarter of all employment falls into one of the four exposure gradients, with significant differences between female and male employment, particularly in the top two exposure gradients (Figure 5). Among male workers, approximately one-fifth of jobs fall within one of the exposure gradients, with 3.1% in gradient 3 and 2.4% in gradient 4, the highest exposure category. In contrast, not only is the total share of female employment exposed to GenAI notably higher, but the differenc e is also concentrated in the top two gradients, w ith 5.7% of female employment in gradient 3 and another 4.7% in gradient 4. These disparities are even more pronounced in high-income countries. Income-based differences in exposure across country groups are also striking, with high-income countries showing the highest share of employment within one of the four exposure gradients (34%). The total share of exposed employment declines significantly as income levels decrease, reaching just 11% in low-income countries. Lower income levels also reduce sex disparities in exposure, primarily due to the lower concentration of occupations in the highest exposure gradients (gradients 3 and 4). Sex disparities are also more pronounced in wealthier regions, particularl y in Europe and Central Asia, where 39% of female employment falls into one of the four exposure gradients compared to 26% for men. These patterns reflect both occupational s ex segregation and the extent to which GenAI -exposed roles, such as clerical, financial, and customer service jobs, are concentrated in higher-income countries. These new estimates are consistent with those given in 2023, but more refined. The previous version attributed 2.3% of global employment to the “automation” category, one percentage point lower than the 3.3% assigned to gradient 4 in the current analysis. In addition, the 2023 estimates classified 13% of employment as subject to “augmentation” and another

global employment to the “automation” category, one percentage point lower than the 3.3% assigned to gradient 4 in the current analysis. In addition, the 2023 estimates classified 13% of employment as subject to “augmentation” and another 16.2% of employment as “the big unknown”. In comparison, the combined employment share for Gradients 1 -3 at the global level in the current analysis is 20 .5%, which is 9 percentage points lower than combined shares of augmentation and “the big unknown” (29.2%) reported in 2023.

4 Calculations of employment aggregates were conducted by David Bescond, ILO STATISTICS. ILO Brief 8 Figure 5: Global estimates of occupations potentially exposed to GenAI (% of employment by sex)  Conclusions As in 2023, our new estimates reflect “potential exposure” to GenAI, not the actual impact on occupations or employment levels. They represent an upper threshold of the percentage of employment that could be affected if GenAI technology were fully implemented. Infrastructure constraints (electricity, broadband), insufficient digital skills, the cost of technology, and inherent operational difficulties associated with the technology are just some of the barriers to full adoption. These calculations also do not account for new jobs that may be created, nor the technological advances that could potentially automate more tasks in the future. As such, th ey present a static snapshot of the exposure of existing occupations to GenAI at the beginning of 2025. The study suggests that few jobs are currently at high risk of full automation. Gradient 4, the category at highest risk, comprises 3.3% of global employment. The occupations within this category have an average score of 0.7, indicating that even in this category there is a small buffer against job displacement, hopefully mitigating the fallout from “technological unemployment”. Gradients 1-3 comprise occupations where the automation of specific tasks is more limited, and there is moderate to substantial variability of task-level scores exists. As such, the “bundle of tasks” that constitutes occupations acts as a buffer

unemployment”. Gradients 1-3 comprise occupations where the automation of specific tasks is more limited, and there is moderate to substantial variability of task-level scores exists. As such, the “bundle of tasks” that constitutes occupations acts as a buffer against full automation. However, this does not mean that demand for these occupations will remain stable. If efforts are not made to help them evolve with GenAI and integrate new tasks, even partial automation of existing tasks could lead to a decline in overall job demand in occupations found in higher exposure gradients. ILO Brief 9 More significantly, the findings suggest that the greatest effect of generative AI on occupations is in transforming work. Integration of GenAI in the work context implies changes in occupational roles, with potentially important implications for job quality. If some tasks are automated to allow workers greater time for more fulfilling work or to heighten their expertise with the help of AI tools (Autor, 2024), this could be positive for job quality. However, if the technology is used to standardize work processes and reduce human autonomy, if it is applied with the sole purpose of increasing monitoring, or if it is not well designed nor well integrated into the workplace, job quality might suffer. For this reason, social dialogue and workplace consultation are needed to ensure that the development and integration of GenAI tools at the workplace is a boon for both working conditions and productivity.

Disclaimer: The views expressed herein are those of the authors and do not necessarily reflect the views of the International Labour Organization.

Contact details International Labour Office Route des Morillons 4 CH-1211 Geneva 22

Switzerland T: +41 22 799 8481

E: gmyrek@ilo.org https://doNQ9406