Global corporate practice has encountered a phenomenon that has already earned the name "AI boomerang": companies are mass-laying off staff, handing tasks over to neural networks, and then inviting the dismissed employees back. In Russia, this scenario is practically unseen — and the reasons lie not in the quality of the technology, but in the structure of the labor market.

Over the current year, Microsoft, Barclays, the Kroger supermarket chain, and dozens of other corporations have announced layoffs tied to the adoption of artificial intelligence. In some cases, positions were deemed redundant; in others, budgets were freed up for technological development. Tens of thousands of people lost their jobs. However, a significant share of these stories had an unexpected ending: companies fairly quickly discovered that neural networks and AI agents handle tasks worse than expected and began bringing people back into the same positions.

The most telling cases

  • Swedish installment-payment service Klarna is once again hiring live support operators. Company head Sebastian Siemiatkowski admitted that the bet on chatbots in the name of cost savings proved excessive.
  • Ford in June hired or promoted 350 experienced engineers after AI failed at quality control.
  • Meta refused to replace employees with AI agents: people were spending too much time fixing mistakes made by the machines.

How widespread this reversal has become is hard to judge — companies are in no hurry to publicly admit that their bet on AI lost. As Lucy Beaumont, senior vice president at British HR company SHL, notes, shareholders like it when a business cuts costs, but admitting that it rushed into AI has to be done without extra noise.

Why neural networks couldn't handle it

Recruiting company Robert Half surveyed hiring managers in finance, marketing, law, and medicine — many described the same picture. First the company eliminated positions because AI was supposed to do their work, and then it reopened them. The reasons are similar everywhere: neural networks had to be monitored far more closely than planned, and AI doesn't know how things actually work inside a company. That knowledge accumulates in living employees over years and isn't written down anywhere.

Betsy Summers, an analyst at research firm Forrester, calls this "invisible work": people notice a colleague is out of sorts, gather to discuss a problem without any schedule, go the extra mile when a task grabs them. None of this is in the job description, but it is precisely what a team rests on. When such people are removed and AI agents are expected to deliver the same result, companies usually miscalculate.

Experts advise those now being called back not to return on the old terms. Brooks Holtom, a professor of management at Georgetown University, recommends asking for a one-year contract with compensation in case of early dismissal. According to Beaumont's observations, returning employees demand more money and a direct answer to the question of what the company plans to do with its staff going forward.

The Russian scenario: plugging gaps, not cutting costs

Russian business is also eliminating positions with the help of AI, and the most notable example is Sberbank. The bank abolished 65% of manager positions responsible for regional branches, and in 2025 it cut its central office by 20%. Herman Gref, at the Moscow Startup Summit, admitted that he isn't worried about his own chair — he is more concerned about what will happen to 200,000 employees. According to the head of Sber, he spends a lot of time searching for new lines of business where people freed up by automation can be transferred.

At the government level, layoffs are explained differently than in the West. Mikhail Kirsanov, director of the employment department at the Ministry of Labor, stated that AI and robots will help close the worker shortage in a number of industries. Experts at the Presidential Academy calculated that by 2035 the economy will need 69 million workers instead of the current 76.7 million. That is, where Western companies cut for efficiency, Russian ones plug gaps that simply no one is left to fill.

The scale of adoption is quite comparable: more than 70% of large Russian enterprises already use generative AI in at least one work process, and almost half of employers use neural networks in hiring. According to HSE calculations, AI will affect about a third of jobs in the country to one degree or another, and one in ten falls into the high-risk zone. But less than 1% of positions can be fully automated. The study's author, Rostislav Kapelyushnikov, believes that neural networks will not destroy professions but will redistribute tasks between humans and machines.

So far, the main consequence for workers in Russia has not been a wave of layoffs but a skills race. Sberbank's Center for Macroeconomic Research analyzed 40 million job postings and found that in two years the share of offers requiring proficiency in generative AI grew fivefold. Demand has long since moved beyond IT and reached marketing, design, and administration. Vacancies with a large volume of routine have become fewer — their share fell from 30.5% to 24.1%. Center head Alexander Isakov links this to the labor shortage: technology raises productivity and smooths out the lack of people, but at the same time stratifies the market.

There have been no loud "AI boomerang" stories in Russia yet. It is hardly because neural networks work better here. Rather, with a shortage of personnel, companies have no spare people whom they could first fire and then call back. But the technology's limitations are the same everywhere, and Russian business will need human oversight just as much as Western business. The winners will be those who master neural networks before their employer decides it can manage without them.

Cryptalist analysis: The "AI boomerang" case demonstrates a fundamental divergence between market expectations and the actual performance of generative models — capital overestimated the speed of replacing cognitive labor, while corporations underestimated the cost of "invisible" knowledge. For investors, this is a signal: short-term growth stories around automation should be discounted, since the adoption cycle will prove longer and more expensive than advertised. The Russian labor market, meanwhile, is moving along a different trajectory — the personnel shortage turns AI from a tool of reduction into a tool of compensation, which shifts the economic effect from cost optimization to skills redistribution.