Companies are returning employees who were laid off due to AI. What you need to know
Large companies, contrary to forecasts, began to re-hire employees who had been reduced due to the introduction of AI tools. If in the last year and a half, employers in the United States believed that fewer staff would ensure their rapid growth, now the situation has changed dramatically. Why AI could not replace humans and how modern technologies are changing the labor market — in the Izvestia material.
The Invisible AI Revolution
• Despite promises that the development of artificial intelligence would not affect the number of jobs, some companies secretly carried out large-scale layoffs to free up funds to implement AI tools and speed up production processes. But the expectation that artificial intelligence models would be able to replace real employees did not materialize.
Tens of thousands of corporate employees were secretly cut by Google (Alphabet) and Meta (recognized as extremist and banned in Russia): now both tech giants are quietly recruiting staff back. IBM reduced its staff as a result of the aggressive implementation of the automated internal personnel management system AskHR and as a result faced serious delays in solving problems. After shifting its focus to automated systems, the Shopify e-commerce platform faced a lack of control and expertise in the field of AI and was forced to rethink its HR strategy.
• An international study of Orgvue's personnel planning platform showed that by April 2025, 39% of companies had reduced their staff in order to implement AI, and more than half of them (55%) regretted this decision. Due to the fact that management has overestimated the advantages of artificial intelligence, companies are hastily returning former employees and reopening vacancies. Foreign analysts call this effect an AI boomerang.
Disadvantages of AI
• The reassessment is related to several vulnerabilities of artificial intelligence, due to which the work with its tools must be carefully checked by humans. The AI is hallucinating. A chatbot on the Air Canada website told the user that the airline would allegedly compensate for part of the flight costs in case of loss of a loved one. After that, he decided to buy tickets to fly to his grandmother's funeral. In fact, the company did not have such a service, but in 2024 the court ordered Air Canada to pay compensation to the client, since the company is responsible for all information posted on the site, including the chatbot.
• The AI is wrong. Ford, which cut hundreds of employees in the hope that the introduction of artificial intelligence systems would allow it to cheaply create a reliable product, faced a sharp drop in quality and forced recalls due to transmission defects and short circuits in the engine compartment revealed during operation. The company paid for this experience with loss of reputation and high costs — in 2024 alone, $4.8 billion was spent on warranty repairs. From 2023 to 2026, Ford brought back most of its former employees, who were supposed to fix AI errors and train young engineers.
• AI does not know how to admit mistakes. This was discovered during testing of the system using automated bots for car cafe ticket offices. As part of the partnership agreement between IBM and McDonald's, more than a hundred establishments in the United States introduced such a service in 2021. It turned out that the AI assistant does not understand accents, repeats orders incorrectly and refuses to correct them. McDonald's discontinued the experiment in June 2024.
• AI is vulnerable to hacking. The British express parcel delivery service DPD implemented an AI chatbot, whose functions included tracking parcels, processing complaints and forwarding requests. A disgruntled customer of the company hacked the bot and added new features to it: criticism of DPD itself, recommendation of competitors' services, and insults against the company's customers. After information about this spread on the Internet, the company disabled the chatbot.
• AI is incapable of human communication. In 2023, the Swedish payment service Klarna, whose name has become synonymous with the installment payment mechanism on the Internet, announced that an AI chatbot employs 700 employees and reduced its staff. As a result, customer satisfaction with the services plummeted: problems that required an individual approach remained unresolved. By 2025, the company had imperceptibly restored the number of employees, returning some of the dismissed employees.
• AI is expensive. Previously, AI assistant services for subscription fees were very cheap because the field was subsidized, but now AI system providers are introducing a "token fee" (a piece of data processed each time an AI agent is contacted). As a result, an artificial assistant may be more expensive than a real employee. In May 2026, it was announced that the AI assistant for GitHub programmers, Copilot, would switch to a token payment system: monthly developer costs could rise from $28.12 to $746.01. At the same time, many companies have already laid off staff in order to finance the introduction of AI to reduce costs.
AI is changing the way we work
• The business is not going to abandon AI, but at the same time the need for real employees will only grow. An artificial assistant can successfully solve individual routine tasks, but must work under human supervision. In an Orgvue study in 2025, 80% of business leaders said they plan to train employees to use AI tools more effectively.
• You will not be able to save money on the implementation of AI. Companies will not only have to set a budget for employee training, but also increase salaries to attract employees with in-demand skills working with AI tools.
• There are areas where you cannot rely on AI. In particular, using an AI model as a copywriter makes content less effective at attracting customers, and in the financial sector, AI can conduct transactions, but is unable to analyze anomalies, as an experienced accountant does. Foreign experts have identified a "60/40 gap" in the work of AI assistants: artificial intelligence can successfully perform about 60% of repetitive work processes, but is absolutely unsuitable for the remaining 40% of tasks that require skills in analyzing complex data, building customer relationships, and strict quality control.
Переведено сервисом «Яндекс Переводчик»