Contents
Artificial intelligence (AI) has become one of the most important technological advances (link in Russian) of recent years, capable of significantly altering the economy and the labour market. Yet, the way AI is and will be doing so remains a topic of discussion (1, 2) (links in Russian). On the one hand, AI is expected to accelerate productivity growth by automating routine operations and expanding employees’ capabilities. On the other hand, there are concerns that, as a new form of automation, AI may replace human workers in production, thus causing unprecedentedly high unemployment, lower living standards for a considerable part of the population, and a sharp rise in inequality.
If the previous waves of technological progress mostly affected low- and medium-skilled workers, AI is, by contrast, expected to crowd out the most qualified groups of employees from the labour market. These are the concerns that fuel alarmist forecasts regarding the future employment catastrophe due to the rapid diffusion of AI. Our analysis (link in Russian) suggests that such predictions are largely unwarranted with respect to the Russian labour market.
Global picture
Researchers use two approaches to assess AI’s impact on the labour market. The first analyses the actual parameters of the use of the technology, i.e. how extensively companies use AI, which tasks it is performing today, and how it affects productivity.
The second assesses the potential impact, i.e. the share of jobs that may be fully or partially automated owing to the widespread adoption of AI. This indicator – AI occupational exposure – helps in grasping the scale of potential problems and in understanding whether the government should prepare to address them in advance. It can be calculated for the economy as a whole and for individual occupations or socio-demographic groups.
It is worth noting that even a strong potential impact of AI on an occupation does not per se suggest that it will soon be eliminated due to automation. Everything depends on the relationship between AI-targeted tasks and those falling within the responsibility of human workers, namely whether AI can perform them or whether it can augment the human performance of these tasks. In the second scenario, where employees’ skills are complementary to AI, demand for their labour may even rise. This means that high (although not extremely high) levels of AI occupational exposure should not be regarded as an alarming sign or as a portent of its impending and inevitable demise.
According to a number of studies, the estimate of AI occupational exposure for many countries hovers around 20–30% (1, 2, 3, 4, 5, 6).
As regards the estimates for various socio-demographic groups, the picture is fairly consistent around the globe (1, 2). AI-related risks are higher:
- for women than for men;
- for the young than for the elderly;
- for urban residents than for rural residents;
- for employees with a degree than for those with lower education;
- for more skilled workers than for less skilled ones;
- for the self-employed than for the employed;
- for those earning higher wages than for those with lower wages; and
- for employees engaged in IT, finance, business services, or public administration than for those working in agriculture, mining and quarrying, or construction.
In addition, the potential impact of AI on employment is typically substantially higher for developed countries as compared to developing ones. Given that Russia is among the countries with medium to high incomes, it can be expected to experience less severe turbulence resulting from the accelerated diffusion of AI through the economy compared with the most technologically and economically advanced countries.
Risk assessment
In my study, I use a scoring system developed by a group of researchers from the International Labour Organization (ILO) based on the International Standard Classification of Occupations (ISCO-2008). Their analysis compares tasks within various occupations with the current capabilities of AI. When AI starts to perform a task previously done by human employees, that task is highly likely to be automated in the future. The next step is to shift from individual tasks to entire occupations: if AI can perform most of the tasks within a given occupation, the latter is considered highly exposed to automation.
Globally, according to ILO-based estimates, around 25% of employees face a material risk from AI adoption, from low to very high. In developed countries, this level reaches 34%, with 17% at the highest AI exposure. In low-income countries, it stands at only 11%.
The empirical analysis of the Russian labour market is based on microdata from Rosstat’s Labour Force Survey. Since the Russian National Classifier of Occupations 2014 (link in Russian), used in the survey, is harmonised with ISCO-2008, this made it possible to adapt AI exposure indices for various occupations to the Russian data and, for the first time in the Russian literature, measure the scale of the potential impact of AI on the Russian labour market.
The study method is based on the detailed classification of about 2,500 occupations and 30,000 work tasks. Following data aggregation, the list of aggregated tasks decreases to 3,000, and the list of occupations decreases to 423.
Each task is assessed using an automation risk scale from 0 to 100, where 0 indicates that it is inherently not amenable to automation and 100 indicates that it can be fully automated using AI. Having consolidated the estimates for all the tasks comprising a given occupation, I derive an integral index of AI’s potential impact for the entire occupations, with 0 denoting that the job can be done only by human workers and 1 denoting that the job can be done by AI without any human involvement.
Based on the indices obtained, the level of AI exposure was disaggregated into six categories: 1) very high; 2) significant; 3) moderate; 4) low; 5) minimum; and 6) zero.
To analyse the Russian data, the first four groups are merged into a consolidated ‘material risk’ category. These are groups with AI exposure risk that cannot be considered negligible.
Areas of AI exposure in the Russian labour market
For all Russian employees, AI occupational exposure comes in at 0.3, that is, around a third of all work tasks in the Russian labour market can theoretically be automated with the help of AI.
For the socio-demographic groups of men and women, as well as urban residents and rural residents, the differences appear to be minor.
By contrast, the risk grows substantially as the education level rises: workers with tertiary education are far more exposed than those without it. The gap between employees with higher and incomplete secondary education stands at 0.37 vs 0.2.
Among occupational groups, white-collar office workers (link in Russian) are the most exposed, with around one in two of them facing the risk of losing their job or, at best, witnessing their job functions transform radically.
Nevertheless, office workers account for less than 3% of all employed people in the Russian economy, therefore, the consequences should be limited even in the worst-case scenario. However, a number of researchers believe that the share of office workers in total employment, calculated based on the Labour Force Survey data, might be underestimated, as it is much higher in all developed countries. If this is true, AI adoption may have more severe implications for the Russian labour market.
The risk is slightly lower, but still very significant, for the upper three white-collar groups, that is, managers and highly- and medium-skilled workers. For them, the index hovers around 0.4.
Contrastingly, all blue-collar occupations related to physical labor, demonstrate very low AI exposure, typically not exceeding 0.2.
The analysis of the sectoral structure shows that AI diffusion is most likely to affect finance, IT, professional services and research, public administration, and trade. For industries other than the services sector, AI does not seem to be a serious threat. In agriculture, for example, the index of AI exposure is only 0.19, indicating that it is almost completely protected from the potential shocks associated with AI adoption.
The breakdown of the main socio-economic groups by level of AI-related risk demonstrates that this risk is zero for around half of all jobs in the Russian economy. Combined with jobs for which risk is minimum, this indicator shows that at least two-thirds of Russian employment falls almost entirely outside the scope of AI’s impact.
Very high AI exposure is equivalent to the risk of full automation and, as a result, the displacement of workers from their jobs. For the economy as a whole and for almost all socio-demographic groups, the share of employees facing very high risk does not even reach 1%.
Only two occupational groups are exceptions: finance, where the risk of full automation threatens one out of every ten workers, and office jobs, where it is real for one in seven employees.
For the economy as a whole, very high and significant risk levels are recorded for approximately one in every ten Russian workers, which is about 1.5 times lower than in developed countries. By contrast, the share of workers exposed to material risk – about one in three people – is on a par with the similar measure for developed countries.
That is, almost a third of all jobs in the Russian economy can be done by AI at least to a certain extent. This means that Russia can be classified as a country with a moderate exposure to AI’s impact on the labour market.
Unwarranted alarmism
The findings suggest that in Russia, AI poses a serious risk only for a rather limited range of occupations, while others do not face this threat at all.
The majority of jobs in the Russian economy are almost fully protected from AI diffusion due to the nature of the work tasks. At the same time, the share of jobs that could be threatened by full automaton if AI takes over all the functions previously performed by human workers is negligible – less than 1% – and even this figure is likely an overestimation. This means that AI adoption in Russia will result not so much in the elimination of entire occupations as in the reconfiguration of tasks as part of the existing jobs, which will definitely remain.
In the foreseeable future, the labour market is more likely to be gradually adapting to the new technologies, rather than developing into a catastrophic scenario, as is predicted by many. In this sense, AI will probably not differ fundamentally from the previous waves of technological progress. Thus, the widespread fears of extremely high technological unemployment that many deem inevitable look like unwarranted alarmism.
In this regard, two points need to be clarified. First, the analysis answers the question of what would happen to the Russian labour market if AI were adopted to the fullest extent, here and now. In reality, AI adoption, as is the case with any other new technology, will look like a dynamic process unfolding over a fairly extended period of time. In other words, the real impact of AI is likely to be much weaker than the potential impact due to the many barriers which will inevitably appear on the path of its diffusion. Consequently, even these moderate estimates appear to overstate the real dangers posed by AI to the Russian labour market. The protracted nature of the process also means that market participants will most probably have enough time to adapt to the new reality.
Second, the analysis relies on today’s AI capabilities. Further technological progress might expand these capabilities substantially, making the negative implications for the labour market much more drastic than this paper shows.
The full text of this reseach is published in the journal Voprosy Ekonomiki, 2026, No. 8.
