We were promised that AI would save time. In practice, for quite a few employees, something slightly different is happening: They still work during the day, but now they also learn at nights and on weekends.

After a full workday, when the computer is already supposed to be shut down, a tutorial on ChatGPT is opened. Then a video on Claude. After that, another attempt to understand what exactly to do with Cursor, AI agents, automations, prompts, updates, new versions, and a new tool that everyone on LinkedIn "must know" by now.

Thus, almost without us noticing, artificial intelligence has created a new race. Not just a race against technology, but a race against other employees, against employers' expectations, against a labor market that broadcasts a clear message: Whoever does not stay updated, gets left behind.

Until recently, professional learning was something done once in a while: A course, a seminar, corporate training, perhaps an additional degree for those who were truly invested. But the AI era has changed the pace. Now, it is not enough to learn a new tool once. You have to keep running.

A survey conducted by EY, the results of which were presented in an article published in Business Insider, shows that 85% of office workers in the US learn to work alongside AI agents outside of working hours. In addition, 83% of office workers say that most of their knowledge about AI was self-taught, and 59% note a lack of sufficient training from the organization as a barrier to developing skills in the field. Meaning, employers want workers who know how to work with AI, but do not always give them the time, the budget, or the framework to learn it properly.

This is exactly where the new "learning tax" was born: The private time that employees invest in order to remain relevant. It does not appear on the payslip, is not always considered training, and is not necessarily defined as part of the role. But in practice, more and more workers feel they have no choice.

It is important to say: AI can indeed save time. A lot of time. BCG's AI at Work 2026 report found that 42% of employees who use AI regularly save at least eight hours a week, nearly an entire workday. This is a dramatic figure, and it explains why organizations are rushing so fast to introduce these tools into almost every department.

But the big question is what happens to the saved time. Does the employee get it back? Is it used for deeper thinking, professional development, creativity, and strategic work? Or is it simply translated into a new expectation to do more, faster, and with fewer people?

In quite a few cases, it seems the answer is closer to the second option. If preparing a presentation once took half a day, today a first version is expected within an hour. If writing a document once required several hours, now it is obvious that one can "let the AI start." And if the tool knows how to summarize, draft, check, translate, and analyze – then why not accomplish yet another task?

Thus, the paradox of the AI era is created: The technology shortens processes, but simultaneously raises the bar of expectations. The work does not always disappear. It simply contracts in time, and then refills with more work.

And there is another part to this race, less glamorous and much less talked about: AI does not work alone. At least not really. It needs someone to write the prompt for it, understand what to ask, check the result, correct mistakes, ensure there are no hallucinations, polish phrasing, and take responsibility for what goes out. In other words, employees are not just using AI. They are also supervising it.

In Glean's Work AI Index 2026 report, this phenomenon was dubbed "botsitting" – supervising the bot. According to the report, employees invest hours every week monitoring, correcting, and improving AI outputs. Instead of performing the entire task themselves, they manage the machine that executes it, and then fix it when it misses the mark.

And this is precisely where part of the fantasy shatters. AI can write text, but someone needs to know if it is correct. It can analyze data, but someone needs to understand if the conclusion is logical. It can suggest code, but someone needs to check that it does not break the system. It can draft an email, but someone needs to ensure it is not embarrassing, erroneous, or legally hazardous.

Meaning, the new skill is not just knowing how to "operate AI." It is knowing how to manage it, question it, correct it, and identify when it takes a shortcut and when it leads into a pitfall.

In the end, the story of AI in the job market is not just about whether it will replace workers or not. This is an important question, but it is no longer the only one. The more immediate question is what it does to the workers who remain.

And the answer is that it changes the pace of their professional lives. It requires them to learn faster, adapt faster, check more, prove more, and above all, not fall asleep on watch. In a world where every week a new tool is released and every month an update arrives that promises to change the rules of the game, the new employee is measured not only by what they know today, but by how fast they are capable of learning the next thing.

On one hand, this is an opportunity. Employees who learn to work correctly with AI can improve performance, save time, open new possibilities, and make themselves more relevant. On the other hand, this is also a clear danger for burnout. Because if every time-saving turns into another task, and every new tool turns into more homework, it is difficult to call this a relief.

Therefore, the question that employers will have to answer in the coming years is not only "Do our employees know how to use AI?", but also: Who gives them time to learn? Who guides them? Who pays for the training? And how do we ensure that the technology meant to ease the work does not turn into another layer of pressure?

Because as of right now, for too many employees, the AI revolution looks less like a shortcut and more like the Rat Race 2.0: Working by day, learning at night, and in the morning returning to the office to discover that the tool they finally understood has already managed to upgrade.