Contents
Economists’ discussions on the impact of generative artificial intelligence (AI) on the labour market are mainly focused on employment, the assessment of job automation risks, and the potential unemployment rate resulting from the technological revolution (1, 2, 3) (links in Russian). The fears are further fuelled by the fact that generative AI is quickly reshaping the perception of the substitution of human labour by technologies.
Earlier stages of automation mainly affected low- or medium-skilled labour focused on performing routine work (both physical and intellectual). Concurrently, the demand for highly skilled specialists involved in non-routine intellectual activities was growing. Today, the new technologies extend to traditionally human-exclusive roles and tasks, which have only recently become exposed to them.
That said, AI is expected to lead to a substantial increase in productivity by complementing human skills. Of the two effects – labour substitution or labour augmentation – the one which eventually prevails will shape AI’s long-term impact on the labour market.
The technology’s influence on employment and incomes is determined by the tasks it automates and the new tasks it generates, rather than the mere fact of its adoption. This relationship determines whether the number of jobs and the returns to specific human skills increase or decrease.
Potentially, AI may automate all the tasks within certain jobs, but it may also change their content and skill requirements without taking over the jobs themselves. Still, to be used effectively, new technologies require a higher level of education and cognitive abilities and therefore often prove to be more helpful for skilled employees. Thus, AI is bringing about job segmentation rather than job substitution. This means that less skilled specialists whose tasks are easily performed by AI have to leave the market while more experienced employees pivot away from direct competition with AI, delegating routine work to it and focusing on more complex and challenging tasks.
For this reason, young specialists that are early in their careers have faced the most severe negative consequences from the onset of generative AI adoption. People in this cohort have vast general knowledge but lack professional expertise, and they are thus more exposed to AI substitution than AI augmentation.
However, in addition to disputes over the future of jobs, another equally important issue arises: what is happening to the incomes of those already using AI? If the technology actually makes a difference to labour productivity, it is bound to influence wages.
The wage ‘premium’ for certain skills is one of the key issues in labour economics. The wage is a specific market bellwether, showing real-time demand for competences. Concurrently, it is a measure of individual productivity. An employee who masters a new tool and thus demonstrates better productivity is expected to earn more.
The same logic applies to generative AI: if its use actually boosts productivity, AI users should be paid higher wages. This is particularly the case for positions in which the new technologies are more applicable. Our study (link in Russian) presents the first assessment of the effect of AI use on wages in the Russian labour market.
AI use in Russia
Until recently, most studies on Russia analysed the impact of AI using data collected before the launch of ChatGPT in 2022. The Russian Longitudinal Monitoring Survey (RLMS) of the Higher School of Economics is the first source helping answer the question of who specifically uses generative AI in the Russian labour market and how its use affects wages.
The analysis uses data from 18,800 respondents aged 18 to 65 over 2023–2024, taking into account both the fact of AI use (including for business purposes) and the frequency of use (sporadic or regular use). To separate the effect of an increase in productivity from the effect of the self-selection of more intelligent and highly paid employees as AI users, our models factor in pre-2022 wage levels, employees’ individual characteristics (gender, age, education, etc.), as well as their job parameters (profession, sector, company size, etc.).
The data show that the number of users of generative AI in Russia is so far insignificant, with the share of respondents who have used AI tools at least from time to time averaging only 10%. Only 5% of employees have used AI for work-related purposes, and only 1.5% do so regularly.
These figures are at odds with expectations, as according to earlier studies (link in Russian), Russian enterprises lead the way in AI adoption. Moreover, they are significantly below the figures reported by other countries: by late 2024, around 40% of the US population aged 18–64 had used AI in their work or daily life, with 9% using it regularly.
In Russia, as elsewhere in the world, AI is most frequently used by young and educated employees, presumably skilled, mostly engaged in IT and living in big cities.
AI is most actively used by ICT specialists (almost 40%), as well as by people employed in finance, science, education, the mass media, and law. The number of AI users is significantly lower outside these sectors. This is particularly true of blue-collar workers and low-skilled labour.
Benefits: possible, but not for everyone
Studies show that AI use is strongly linked to higher wages. On average, AI users earn 13% more than those going without it. However, the scale of the effect depends greatly on the frequency of use, with sporadic and regular use entailing wage premiums of 11.6% and as much as 26.9%, respectively.
The effects are even more notable for highly skilled specialists. In this cohort, sporadic AI use can yield a premium of around 18.2%, while regular use can deliver a premium of up to 45.8%, which is almost double the average for all employees who consult AI regularly.
The observed effect is aggregate in nature. It is based on the actual increase in productivity, which is most likely to be seen by regular AI users, and the fact that the technology is most frequently used in positions that were originally highly paid, such as in IT. After adjusting for individual effects and job characteristics, including employees’ pre-2022 wages, a stable wage premium is observed only for regular AI users.
This brings us to the key finding that the value of AI is higher where it is more applicable. In other words, in the early stages of AI adoption, the technology and human capital complement one another. AI delivers the largest benefits to those who are already well-educated and highly skilled.
Currently, generative AI serves as an aid rather than a labour substitute. When used in combination with other components of human capital (education, skills, and experience), the tool can increase output many times. The higher wages of AI users suggest that the use of AI actually drives a measurable increase in productivity, at least for certain categories of highly skilled employees. This increase is achieved through regular AI use to automate routine tasks within existing jobs, rather than through the complete substitution of jobs.
Furthermore, the greater applicability of AI in jobs that are traditionally well-paid is evidence that the technology consolidates existing inequality rather than causing new inequality. Although AI enables less experienced and less skilled specialists to narrow the skills gap, its adoption actually expands this gap rather than helping bridge it. The real value of AI is unlocked primarily by those who already have high incomes and high skills.
A look into the future
AI is undeniably reshaping the economy and the labour market, but this does not imply immediate negative consequences. As the world evolves, skill requirements are transforming. Who will emerge on top and who will be forced to completely reinvent themselves depends on employees' readiness to adapt – sometimes following the market and sometimes anticipating its demands. For skilled employees, this underscores the need for continuous upskilling and the regular adoption of AI tools.
This poses significant challenges for the education and professional training systems. Learning can no longer be limited to one-off courses. Instead, it demands the systematic integration of AI into educational programmes at all levels.
Generative AI has already become a major factor influencing workers' earnings. However, the long-term effects remain difficult to predict. We are currently witnessing the initial stage of the adoption of the technology, the phase in which the wage premium is at its peak, driven by limited adoption among employees.
As AI becomes ubiquitous and the supply in the labour market grows, the premium observed today may diminish or vanish entirely – much like what previously happened with basic digital competencies. In that scenario, proficiency with generative AI will become a standard skill requirement for securing highly paid positions.
Generative AI can indeed serve as a powerful catalyst for productivity and income growth, but only for those who are prepared to use it regularly and purposefully. For everyone else, it will remain either a threat or a missed opportunity.
The full text of this study (link in Russian) is published in the journal Voprosy Ekonomiki, 2026, No. 8.
