Employees who were replaced by AI are smiling. They've found that AI can't handle the work. More and more companies are starting to "call people one by one" to invite them back...

Last year, there was a popular saying in Silicon Valley: "AI will take your job."

From Google and Amazon to various startups, many companies publicly announced "optimizing human resources with AI" and showed investors a blueprint of future efficiency. However, a year later, reality gave them a slap in the face - more and more companies are rehiring the employees they laid off.

Recently, a global report released by the human resources data platform Visier showed that among the 2.4 million employees in 142 companies it tracked, 5.3% of the laid - off employees were eventually rehired by their former employers. Although this figure may seem low, the trend is very obvious - this proportion has been relatively stable since 2018 but has increased significantly in the past two years and is accelerating.

This means that the myth of "AI replacing humans" is facing a real - world shock.

An Unfulfilled "AI Promise"

Andrea Derler, the chief researcher at Visier, described this situation as a "cooling - off period between companies and AI."

She pointed out that when many companies laid off employees last year, they gave the reason that "AI can automate some positions." But in fact, this statement is more of a "convenient excuse for layoffs" rather than being based on real business results.

According to Visier's analysis, after many companies introduced AI systems, they were indeed able to improve efficiency in some processes, such as customer service Q&A, data entry, and the initial generation of project reports. But the real problem is that AI usually can only "take over tasks" rather than entire positions.

That is to say, in a complete position (such as a market analyst, a development engineer, or a project coordinator), AI can only replace some of the operations. The remaining parts that require judgment, coordination, and experience still need humans. As a result, these companies found that "they saved half of the staff but still had to spend more human resources to make AI truly usable."

Moreover, a more realistic problem is "money."

As Andrea Derler said, many executives completely underestimated the cost and complexity of AI implementation when promoting AI strategies: "AI infrastructure is not as simple as connecting a model to an API. It also includes a whole set of systems such as servers, data pipelines, security audits, and model monitoring." - And these expenditures often far exceed the company's initial budget for introducing AI.

Especially in the enterprise - level scenario, AI models need to run in the internal environment, which inevitably involves issues such as privacy, compliance, and security. These factors will cause the cost to soar. Some companies originally planned to offset this cost through layoffs, but soon found that the speed of AI launch could not keep up with the complexity of the business.

So, when the new AI system cannot smoothly take over the old processes after going live, the company has no choice but to re - hire the original business backbones.

MIT: 95% of Enterprises Haven't Made Money from AI

In addition to Visier, a recent survey released by the Massachusetts Institute of Technology (MIT) also confirmed this - about 95% of enterprises globally have not obtained any quantifiable financial returns from AI investment.

In other words, AI has not only failed to become a "productivity" on the balance sheet but has also become a new center of capital expenditure. Steve Sosnick, the chief strategist at Interactive Brokers, was more direct in his evaluation: "It seems that a lot of the huge investment in the AI industry has not been spent effectively."

So, this has put many companies in an embarrassing cycle: they laid off employees and invested in AI; but AI did not save the expected human resources; instead, they had to re - hire people to make the system run. As a result, "cost - cutting with AI" has become "AI backfiring."

Moreover, even from the perspective of traditional cost control, layoffs are not a "sure - win" operation.

According to data from the labor planning platform Orgvue, for every $1 in salary cost a company saves, it actually spends an average of $1.27 on implicit costs such as severance pay, unemployment insurance, re - recruitment, and training.

"Layoffs are never free," Andrea Derler emphasized. "And many companies underestimated the complex chain reactions they bring."

Is the Myth of "AI Replacing Humans" Collapsing?

Looking back at this wave of AI, it has indeed brought about huge industrial changes, but it has also exposed many overlooked realities.

Visier said that this wave of "rehiring" is spreading across various industries, especially in companies that tried to quickly introduce AI tools. For example, some technology companies laid off employees in customer service, operations, or testing positions. Only later did they find that the content generated by AI was unstable and the error rate of work - order recognition was high, and they had to urgently recall experienced employees familiar with the business processes to "remedy" the situation.

"Companies that overestimated the savings brought by AI may eventually have to call one by one to re - hire the employees they laid off."

To some extent, the AI - related layoff wave may just be a part of the companies' trial - and - error process, and this "rehiring" phenomenon shows that they are learning to "correct their expectations" - instead of fantasizing that AI can replace everything, it is better to find a new balance point for the coexistence of AI and humans.

But this reversal from "layoffs" to "rehiring" at least reminds us of a reality: the power of technology always needs to be harnessed by humans.

Reference link: https://www.techspot.com/news/110139-new-data-shows-companies-rehiring-former-employees-ai.html

This article is from the WeChat official account "CSDN". It was compiled by Zheng Liyuan and published by 36Kr with authorization.