Has Artificial Intelligence Replaced Humans? Why Businesses Are Turning Back to Employees

Until recently, artificial intelligence was seen as the nearly perfect employee. It doesn’t get sick, doesn’t take vacation, can work around the clock, respond to thousands of customers at once, analyze documents, write code, and create content. Companies were promised drastic cost reductions, increased productivity, and the ability to operate with significantly fewer employees.

Riding this wave of optimism, some employers began halting hiring, laying off employees, and delegating their tasks to algorithms. However, the first large-scale experiments showed that there is a big difference between a flashy AI demonstration and its full-scale operation within a company.

To use artificial intelligence, businesses need high-quality data, expensive computing power, integration with corporate systems, security controls, and people to verify the results. If the system makes a mistake, the company is still held responsible, and employees are the ones who have to deal with the consequences.

That is precisely why, following the first wave of layoffs, businesses are gradually shifting to a different approach. AI is being reserved for standard and repetitive tasks, while people are being brought back to roles that require responsibility, experience, empathy, and the ability to handle non-standard situations.

UA.News explains how many jobs have already been automated using artificial intelligence, why companies are rethinking their plans for mass layoffs, and whether an algorithm is truly cheaper than a human employee.

AI has already led to tens of thousands of announced layoffs

It is practically impossible to determine exactly how many people worldwide have lost their jobs specifically because of artificial intelligence. Companies do not always explicitly announce that they are replacing employees with algorithms. Layoffs are often explained as restructuring, automation, technological upgrades, or the need to cut costs.

One of the most detailed sets of statistics is maintained by the American firm Challenger, Gray & Christmas, which analyzes employers’ public announcements regarding staff reductions.

According to its data, throughout 2025, U.S. companies cited artificial intelligence as one of the reasons for cutting 54,836 jobs. This accounted for approximately 5% of all layoffs announced in the U.S. that year.

In 2026, the number of such announcements began to rise rapidly. In April alone, employers attributed 21,490 planned layoffs to AI. This accounted for approximately 26% of all layoffs announced in the U.S. that month. In the first four months of the year, artificial intelligence was cited in plans to cut 49,135 jobs, or about 16% of the total layoffs during that period.

In June 2026, AI once again became the main cited reason for layoffs. Over the course of the month, 14,029 jobs were attributed to it, accounting for 31% of all announced layoffs. In total, since the beginning of 2026, companies have cited AI in announcements regarding the elimination of 101,743 positions, or about 23% of all layoffs in the U.S. Since 2023, when Challenger began tracking artificial intelligence separately, it has been cited in plans to cut 173,568 jobs.

However, these figures should be interpreted with caution. They refer to plans announced by employers, not confirmation that each of the 173,000 employees was directly replaced by a chatbot or another AI system. Artificial intelligence is often part of a broader restructuring process that also includes cost-cutting, job relocations, and division closures.

Companies often lay off employees even before AI has proven its effectiveness

One of the main problems is that management may decide to cut staff before the new system demonstrates consistent results.

Businesses see that AI can quickly generate text, answer a standard question, or analyze a document, and extrapolate this result to the work of an entire department. However, a single successful task does not mean that the technology can independently manage an entire workflow.

This is particularly evident in customer support. A chatbot can provide payment status updates, locate a section in a manual, or help change a password. But if a customer has several related issues at once, an account that was mistakenly blocked, or an unusual financial situation, the system often transfers the conversation to a human agent.

In September 2025, Gartner predicted that by 2028, no Fortune 500 company would be able to completely eliminate human customer support. Analysts also expected that half of the organizations that had planned to significantly reduce the number of agents using AI would abandon those plans by 2027.

This means that the reality differs from the idea of “plugging in a bot and laying off an entire department.” In most cases, companies have to retain human staff to handle complex inquiries, ensure quality control, and correct errors in the automated system.

Klarna: AI Performed the Work of 700 Agents, but People Had to Be Reinstated

The most famous example of a failed attempt to minimize human involvement was the Swedish fintech company Klarna.

In February 2024, the company announced that its AI assistant had handled 2.3 million conversations with customers in just the first month and taken on approximately two-thirds of all support inquiries.

According to Klarna’s own estimates, the bot handled a workload equivalent to that of 700 full-time agents. It operated in 23 markets, communicated in more than 35 languages, and the average time to resolve an issue dropped from 11 minutes to less than two. According to the company, the number of repeat inquiries decreased by 25%.

The company also projected that the AI assistant would increase its profits by approximately $40 million in 2024.

At first glance, the experiment seemed to prove that artificial intelligence is already capable of replacing hundreds of employees. Klarna froze hiring, reduced its reliance on external call centers, and actively promoted the idea of a lean company where algorithms handle most of the routine work.

However, management later acknowledged that an excessive focus on cost-cutting had negatively impacted service quality. Klarna CEO Sebastian Siemiatkowski stated that when cost savings become the primary goal, the company ultimately ends up with lower quality.

In 2025, Klarna began hiring again. In September of that year, more than two dozen job openings were posted on the company’s website. Management shifted its focus from replacing employees to using AI to improve products and assist customer service representatives.

In 2026, the company continued to use AI for simple tasks but acknowledged that complex inquiries were quickly routed to human agents. According to Klarna’s new assessment, the system could handle the workload of approximately 850 agents, but customers still needed the option to speak with a live agent.

Consequently, Klarna did not abandon the technology, but it did move away from a virtually staff-free model. AI remained the first line of support, while human agents handled complex cases and ensured service quality.

Half of companies may rehire laid-off agents

In early 2026, Gartner released an even more telling forecast. According to the research firm’s estimate, by 2027, 50% of organizations that laid off customer support staff due to AI will rehire people to perform similar work, although job titles may change.

This forecast is based on the actual results of the technology’s implementation. In a Gartner survey of 321 support service managers, only 20% reported that they had already reduced the number of agents due to AI. About half of the respondents noted that staffing levels remained roughly the same as before.

In other words, at most companies, artificial intelligence has not eliminated an entire department. Instead, it has helped employees handle more inquiries, find information more quickly, and prepare responses.

At the same time, 91% of customer service managers in 2026 felt pressure to implement AI. As a result, companies may have rolled out automation not only because it was cost-effective, but also out of fear of falling behind their competitors.

By 2028, more than half of customer service departments may double their spending on technology, but without a corresponding reduction in personnel costs. In other words, companies will be paying for both AI systems and the people who oversee them and handle complex inquiries.

Why Artificial Intelligence Isn’t Always Cheaper Than a Human Employee

The main argument in favor of automation is cost savings. Companies compare an employee’s salary to the cost of accessing an AI model and conclude that software must be cheaper.

However, this is an incomplete comparison.

To use AI in a real-world business setting, corporate data must first be prepared. Often, the information is stored in different systems, is in an outdated format, is duplicated, or contains errors. Before training an algorithm on this data, the company must verify and structure everything.

Next, integration with CRM systems, payment systems, databases, and internal services is required. Additionally, the business pays for cloud computing, cybersecurity, technical support, audits, licenses, monitoring, and model updates.

Furthermore, AI cannot simply be left unattended. People must verify responses, track errors, and prevent leaks of confidential data.

It is precisely because of rising costs, unclear benefits, and insufficient control that Gartner predicts that more than 40% of projects to create autonomous AI agents will be shut down by the end of 2027.

Previously, Gartner also predicted that at least 30% of generative AI projects would be halted after the testing phase by the end of 2025. Among the main reasons cited were low data quality, insufficient risk management, rising costs, and a lack of clear business value.

By 2030, Gartner predicts that the cost of fully automating a single customer inquiry using generative AI could exceed the cost of a contact center agent. This will be due to expenses related to computing, integration, and quality control.

Therefore, the claim that AI is always cheaper than a human only holds true in simple scenarios and when there is a large volume of repetitive tasks.

Most companies use AI but do not see significant returns

Artificial intelligence has already penetrated nearly every industry. According to a 2025 global study by McKinsey, 88% of the organizations surveyed regularly used AI in at least one area of their operations.

However, only 39% of companies reported any noticeable impact of AI on earnings before interest and taxes (EBIT) at the organization-wide level. Researchers classified only about 6% of respondents as companies that derived significant financial benefits from AI.

The survey covered nearly 2,000 respondents from 105 countries. Its results showed that the gap between testing the technology and generating actual profit remains very wide.

The companies that achieved the best results did not simply grant employees access to a chatbot. They restructured their workflows, determined when AI responses should undergo human review, changed team structures, and appointed individuals responsible for managing the technology.

Other studies cited by Reuters painted an even more modest picture: only about 5–15% of executives reported a noticeable increase in profitability or significant business value from generative AI.

Thus, widespread AI adoption does not yet equate to widespread financial success.

AI can create more work than it eliminates

Automation doesn’t always reduce the workload. Sometimes it simply shifts work from one employee to another.

For example, AI can draft a legal document in a matter of seconds, but a lawyer still needs to verify the facts, case law, and references to legislation. An algorithm can generate program code, but a developer must test it, identify vulnerabilities, and ensure it doesn’t interfere with other systems.

In customer support, a bot handles simple questions, but complex inquiries—after a failed automated conversation—are forwarded to a human agent. The agent then has to deal with an already frustrated customer, an incomplete history of the issue, and the need to correct the system’s incorrect responses.

The experience of the American telecommunications provider Verizon is illustrative. The company uses AI for initial call analysis, retrieving customer information, and routing inquiries. However, its representatives acknowledged that about 40% of customers still want to speak with a human and get frustrated when they can’t reach an agent quickly.

Verizon retained about 2,000 first-line support employees and used AI primarily as an assistant rather than a complete replacement for them.

According to the Zendesk platform, its clients can automate between 50% and 80% of inquiries, but these are mostly standard and predictable questions. Complex cases requiring empathy, negotiation, or accountability remain the responsibility of human agents.

IBM automated part of its HR operations but began hiring in other areas

IBM is another telling example.

The company automated part of its HR operations using the AskHR system. AI took over standard employee inquiries, document processing, and some administrative tasks. According to IBM CEO Arvind Krishna, the system made it possible to replace the work of several hundred HR employees.

However, IBM’s total workforce did not shrink proportionally. The company began hiring more programmers, sales and marketing specialists, and other professionals whose work requires critical thinking and interaction with clients.

This example does not mean that IBM rehired all the laid-off HR employees to their previous positions. It demonstrates something else: automating one type of work can create demand for people in other parts of the company.

IBM’s revenue from contracts related to generative AI reached approximately $6 billion, but to fulfill these contracts, the company needs developers, consultants, salespeople, and integration specialists. The technology has reduced some administrative work, but at the same time, it has created new tasks for people.

Artificial intelligence replaces specific tasks, not people entirely

The most realistic conclusion is that AI does not replace entire professions, but rather specific tasks within them.

It performs well on tasks that have clear rules, are repeated many times, and are easily verifiable. These include sorting requests, drafting documents, searching for information, translating standard texts, and answering typical questions.

However, most people’s work consists of more than just these tasks. Employees also interact with colleagues, take responsibility for decisions, assess risks, respond to unexpected circumstances, and consider context that may not be present in a database.

It is precisely for these tasks that modern AI most often requires human oversight.

Even in companies that are actively downsizing, people remain responsible for the final outcome. An algorithm may suggest a solution, but it will not be held accountable to a client, a court, a regulator, or management if that solution turns out to be incorrect.

So, has AI been able to replace humans?

Artificial intelligence has already impacted the labor market and has indeed become one of the reasons for layoffs. In the U.S. alone, starting in 2023, companies have cited AI in plans to eliminate more than 173,000 jobs.

However, these figures do not yet prove that algorithms are capable of completely replacing workers.

Klarna’s experience has shown that while a company can automate millions of routine inquiries, it is ultimately forced to bring people back due to a decline in quality. Gartner forecasts indicate that half of the companies that laid off customer service representatives due to AI may end up rehiring employees for similar tasks.

In addition, more than 40% of autonomous AI projects are at risk of being shut down due to high costs, insufficient oversight, and a lack of clear benefits. More than half of customer support departments may double their technology costs without achieving an equivalent reduction in personnel costs.

Therefore, the main change lies not in the disappearance of humans, but in the redistribution of work between employees and machines.

AI handles speed, scale, and repetitive tasks. Humans retain responsibility, oversight, complex decision-making, creativity, negotiation, and handling non-standard situations.

As a result, the most effective model is not “artificial intelligence instead of humans,” but rather “humans with artificial intelligence.” Companies that have tried to completely remove employees from the process are gradually returning to precisely this balance.

Read us on Telegram and Sends