For some office workers, the start of each month brings a ChatGPT Plus or Claude Pro charge on their credit card statement. Not a corporate card. Not company wellness points. Just money leaving their own bank account. In a survey of 1,000 South Korean office workers conducted jointly by Deel and Remember, a Korean business-card and professional-networking app, 53% of respondents said they pay for premium AI services out of their own pockets. That's more than half.

Some will read that number as enthusiasm; others will read it as proof that workers already know their companies won't wait for them. When a monthly AI subscription sits in the same budget line as an English-conversation app or an exam-prep course, it means this skill has come to be seen as a personal asset — one you take with you when you change jobs, one that belongs to no organization.

Numbers on Top of Numbers, Action on Top of Action

In the same survey, 82.4% of respondents named AI skills as a core element of career growth. Close behind, 89.4% said they are currently doing something concrete to build those skills. Not intentions — actions. The roughly seven-percentage-point gap between the two figures is subtly telling. That's where you find the people who believe it matters but haven't yet turned that belief into action.

The most popular way to learn was simply using AI on the job, at 38.8% — far ahead of taking courses or earning certifications. It's hard to read that figure as mere convenience. It suggests people have already concluded that, when it comes to keeping up with fast-changing tools, hands-on practice beats curriculum-based study. Attaching AI to the report due today or the proposal due this week, and seeing for yourself whether it works, becomes the learning itself.

Read in this context, the 53% figure takes on a different meaning. These are people who felt firsthand that corporate training and company AI adoption weren't keeping pace with their own drive to learn — and decided to stop waiting and pay for it themselves.

The survey also surfaced worries about over-reliance on AI. Notably, this anxiety tends to show up more sharply among frequent users than among people who haven't tried AI at all. The more convenient the tool becomes, the more often the question surfaces: what happens to me without it?

Using AI and Leaning on AI Point in Different Directions

Put two people with the same subscription side by side. One types "draft this email for me" and sends whatever comes out. The other takes the draft, finds the sentences that miss their intent, revises them, and uses that experience to make the next request more precise. They pay the same subscription fee. Six months later, what each has to show for it is entirely different.

Use AI purely as an output generator, and the quality of its output becomes your standard. As the cycle of picking and polishing repeats, your judgment doesn't sharpen — your standards quietly adjust to fit whatever the model produces. The direction is backwards. But someone who uses AI to refine their thinking, stress-test their judgment, or map the terrain of an unfamiliar field finds their thinking growing denser the more they use it. The difference comes not from the tool's features but from the attitude you bring to the tool.

This is where the worry about over-reliance becomes meaningful. As a tool gets more convenient, the muscles you once used without it quietly weaken. The time spent wrestling with a blank first sentence. The process of digging through sources and fishing out patterns yourself. The experience of trusting your own judgment and pushing forward through uncertainty. When AI takes these over, it's hard to tell in the moment whether your abilities are expanding or something is slowly atrophying.

And here the content of the word "skill" becomes the problem. Whether "being good with AI" means quickly extracting the output you want from the tool, or sharpening your own thinking through the interaction, completely changes how you would go about building that skill. With the former, every advance in the tools leaves you scrambling to catch up. With the latter, every advance in the tools makes you sturdier.

There is a view that as technology absorbs repetitive, draining work, we can redirect our energy toward thinking more deeply, judging more accurately, and deciding more deliberately. That happens when the tool works not to replace you but to express you better. And it requires settling your attitude toward the tool first. Leave the direction up to convenience, and the current naturally drifts toward dependence rather than growth.

What to Do First When There's No Corporate Training

Solo entrepreneurs, freelancers, and people on small teams face this problem more directly. There's no in-house AI training program, no tool-adoption budget, no AI task force. People in this category likely account for no small share of that 53% paying out of pocket.

The first move here is not drawing up a systematic study plan. It's pulling out the list of things you actually have to do this week and looking for the places where AI can attach. The starting point for learning isn't an abstract skill-development goal — it's wherever friction shows up in your real workflow.

If there's a draft format you write over and over, you can have AI lay out the structure first while you concentrate on judging and revising. That raises efficiency, and the repeated act of deciding what to keep and what to fix also sharpens your standards. When you need to quickly grasp unfamiliar contract terms or a new market, ask AI for the lay of the land first, then second-guess its answers and use that doubt to refine the questions themselves. Walking into a meeting with an expert already knowing what to ask yields a very different density of insight than walking in cold. And if you often have to make decisions alone, asking AI for a counterargument before committing is a practical move — a structure that fills in for the review step you'd otherwise skip for lack of a second pair of eyes.

None of these three is about getting output faster. They are ways of attaching AI to the process of checking and extending your own thinking. Used in this direction, the subscription fee buys capability, not dependence.

In an era when 82% call AI skills central to their careers, failing to define for yourself what that skill actually consists of means the numbers can fool you. Using a tool a lot and using a tool to grow yourself point in different directions. Which set of muscles the AI app you're opening right now makes you use — the difference already begins with that distinction.