AI as a New Productivity Disruptor: Opportunities and Risks

Artificial intelligence is becoming a new disruptor of labor productivity worldwide

Every major technological wave has changed not only how we work but also whose work becomes redundant. Industrialization and the automation of repetitive operations sharply reduced the demand for blue-collar jobs in plants and factories. The current AI wave affects a different layer of employment: administrative and office work, which until recently was almost impossible to automate. The next stage will be the combination of AI with humanoid robots, which will deliver an additional productivity gain wherever manual labor still remains in production.

The first signs of this shift are already visible in the data. According to the IMF, AI will affect about 40% of jobs worldwide and about 60% in advanced economies. A Stanford University study based on payroll data for millions of American workers shows that employment of young professionals aged 22–25 in the occupations most exposed to AI (software development, customer support) is already 19% lower than it would have been had it grown at the pace of their peers in less affected occupations. Notably, this is happening not through layoffs but through reduced hiring: companies are simply ceasing to take on entry-level staff. There are no mass job cuts across the economy yet, but in my estimate they will become noticeable within 5–10 years. The period is still too short to measure the effect of AI in the statistics: with general-purpose technologies, the returns first go into restructuring processes and training people, and productivity growth becomes visible with a lag, as it did with electricity and computers.

I see the main macroeconomic risk in demand. AI lowers labor costs and increases the supply of goods and services, while the incomes that pay for demand are redistributed from workers to the owners of capital and technology. History has already seen this scenario. In the 1920s, electrification and the assembly line sharply raised productivity, but incomes were distributed ever more unevenly, and demand increasingly rested on credit. When the debt burden became unsustainable, a crisis of borrowers began, falling prices increased the real burden of debt, and a wave of bank failures turned the downturn into the Great Depression. In 1930 Keynes put first among the causes that technical efficiency was growing faster than the economy could absorb the labor it released, and coined the term "technological unemployment" for it. Later, in The General Theory, he made the case for stimulating aggregate demand through the state. That is why the question of how to sustain consumer demand against the backdrop of shrinking labor inputs needs to be asked now.

The process will be extremely uneven across countries. Those that lag in adopting innovation and AI risk losing in global competition, falling into technological dependence on the leaders and growing poorer. In this sense, I consider the President's course of introducing AI as part of state policy strategically right.

However, this course has a flip side, and it is important to spell it out in advance. Kazakhstan's pension system and compulsory social health insurance (CSHI) are built on contributions from wages, and personal income tax (PIT) is one of the main sources of revenue for local budgets. If labor's share of income in the economy begins to fall, this base will narrow along with it. Therefore, alongside the adoption of AI, we need to work out in advance a gradual shift of the tax burden from labor to capital and rents, mass retraining of workers, and the preparation of young people, who should enter the labor market not only with a diploma but also with AI skills.

We at Solva Group are also actively adopting AI. Experience shows that a company becomes AI-native only when its chief executive is an active AI user. I use AI every day, and at the same time I understand that I am tapping barely 1% of its capabilities.

AI is also changing the competitive environment. Small and medium-sized businesses are gaining capabilities that were previously available only to large companies: building a website, automating processes, answering emails, running social media, writing small programs, doing design, promoting products, and collecting and analyzing data from open sources. All of this can now be done independently and for little money. Large companies that do not start to change will find that their internal bureaucracy becomes a barrier to growth, and they may lose out to faster players. At the same time, large players retain a serious advantage in their own data and access to customers, so the winners will be those who can combine scale with speed.

Everyday practice is changing too. We increasingly turn to AI rather than to a search engine with our questions, and this trend will continue. I switched to Android because it is more deeply integrated with AI: Gemini already manages my meetings, alarms and tasks. Sorting and parsing documents, searching through them and answering email are also handled by AI.

Now I build websites and dashboards myself, set up integrations with 1C and other systems, and track metrics directly, without purchased BI systems such as Qlik or Power BI. MCP connectors make it possible not just to obtain data but also to process it through AI agents and immediately get the information I need in the form I need. Such areas are multiplying. Thanks to Telegram's open API, an AI agent can be connected to it, and now the correspondence in my work groups is analyzed by a Claude-based agent, which answers colleagues' routine questions on my behalf, while I make the substantive decisions myself. For multitasking this is very convenient. Kairat Akhmetov, one of the most advanced vibe-coding practitioners I know, helped me figure this out. His level is so high that in his work groups his personal agent communicates with the agents of his employees.

Now I am trying to write more complex programs, mostly for my own development. I already know quite a few people who build entire products single-handedly and are starting to earn money from them. I think that within the next couple of years we will see the first solo unicorns.

However, this transformation has a downside, and the risks here are significant.

The first risk, which I have already written about, is a demand crisis.

The second risk is a large-scale redistribution of markets and the bankruptcy of many companies whose products will be replaced by AI.

The third risk, a long-term one, is the loss of skills. The more knowledge we delegate to AI, the less we learn to do things ourselves, with our hands and our heads. Already now, when vibe coding, one often accepts the AI's suggestion without reading it closely, because one considers it more competent. This is how skills are gradually lost and dependence on AI grows. Over a horizon of centuries, this could lead to degradation.

The fourth risk, inequality, is more complex than it seems at first glance. Within a profession, AI, on the contrary, levels opportunities: in a study of more than 5,000 customer support employees, an AI assistant raised productivity by an average of 14%, and by 34% for novices and the least skilled workers, with almost no effect on the results of experienced ones. A paradox emerges: AI helps beginners the most, yet they are precisely the ones companies are ceasing to hire. Between those who master AI and those who do not, the gap will widen: the former will become more productive and earn more, while the latter will find themselves in a vulnerable position. And the largest gap, as the calculations of Nobel laureate Daron Acemoglu show, will emerge between the incomes of capital and labor.

Ultimately, the question is no longer whether to adopt AI but who will capture the gains from higher productivity. If they go to a narrow circle of technology owners, we will get both a demand crisis and social tension. If we manage to distribute them more widely through education, retraining and adaptation of the tax system, AI will become the main source of income growth and quality of life in the coming decades.

I would be interested to hear how you use AI in your life and work.