Future of work
AI is not only “taking jobs.” It is taking pieces of jobs: the first draft, the first grade, the first search, and the boring copy-paste middle.
At Creverse, the future of work does not arrive as a robot walking into the classroom. It arrives as a grading system called Hummingbird: feedback in seconds, more consistent comments, and fewer humans needed for the first round of marking.
That is the real AI story. Work gets sliced into tasks. Some tasks become cheaper and faster. Other tasks become more valuable because humans still need to judge, motivate, comfort, explain, and notice what a machine misses.
The class notes describe hagwons as learning centers for everything from taekwondo to coding and English. Creverse sits inside that world, teaching English, coding, and math across a school year that runs almost all year long.
The comparison from class is powerful: photography did not kill painting; it forced painting to evolve. AI may do the same to education. If a machine can grade fast, the teacher's new job is not to compete with speed. The new job is to make feedback mean something.
This article asks the bigger question: what happens when that same pattern reaches every workplace?
Creverse is interesting because it makes AI feel local, not abstract. It is not a Silicon Valley demo. It is a Seoul education company using AI to change the daily rhythm of homework, feedback, and tutoring.
The Hummingbird grading system in the notes has three important claims: operational costs cut by 80 percent, turnaround reduced from three to five days to about 30 seconds, and more consistent feedback quality. That is not just a cool feature. That is a redesign of labor.
Before AI, a tutor might spend time checking grammar, writing repeated comments, sorting mistakes, and returning work days later. With AI, the first pass becomes instant. The human can step in later to explain patterns, deal with motivation, or help a student understand why a sentence works.
The danger is obvious too. If schools only care about cost, AI can become a reason to remove support. If schools care about learning, AI can become a reason to give students feedback more often and make teachers more focused on coaching.
Anthropic's Economic Index describes how people use Claude across occupations, regions, and task types. The strongest pattern is that AI is useful where work is made of language: explaining, summarizing, coding, writing, translating, researching, planning, and giving feedback.
That explains why students, teachers, programmers, analysts, and office workers reach for Claude. They are not always asking it to replace the whole job. Often they ask it to unblock a small part of the job.
Visual index for the article, based on public categories and examples from Anthropic's Economic Index. It is not a raw share-of-all-users chart.
The Anthropic geography research says Claude use is uneven. The U.S. leads in total use. India comes next, followed by Brazil, Japan, and South Korea with similar shares. Inside the United States, high-income and knowledge-work regions stand out.
Hover over the tile map. It is styled like the screenshot you shared, but cleaner and easier to use inside the article. Darker states represent higher relative Claude usage in the story's visual index.
Ordering food with screens already changes entry-level service work. AI adds another layer: recommendations, translation, and automated support.
In Seoul, language apps make daily life easier for foreigners. They do not remove the need to learn Korean, but they lower the fear of being stuck.
AI feedback can be useful when it helps students revise quickly. It becomes dangerous when it makes students stop thinking.
Will Robots Take My Job is useful because it turns fear into a research question. The exact percentages should not be treated as destiny, but they help students compare which jobs contain automatable tasks and which depend strongly on human trust.
| Job | AI pressure | What AI does first | Human strength |
|---|---|---|---|
| Cashier | checkout, inventory, payment | customer care, conflict handling | |
| Programmer | boilerplate code, debugging help | architecture, taste, responsibility | |
| Teacher | grading, quizzes, lesson drafts | motivation, trust, classroom judgment | |
| Psychologist | journaling prompts, resources | empathy, ethics, diagnosis, safety | |
| Music producer | loops, rough mixes, idea generation | taste, emotion, cultural timing | |
| Lawyer | document review, first drafts | strategy, accountability, persuasion |
Risk bars are editorial estimates for classroom discussion inspired by automation-risk databases, not guaranteed job predictions.
Students often need someone to notice frustration, not only errors.
People decide what matters, what is fair, and when a result feels wrong.
Humans create meaning from experience, culture, and relationships.
AI can produce a first answer in seconds.
One system can review thousands of short tasks.
AI is good at sorting large amounts of text and data.
AI finds common errors; the teacher explains the thinking.
AI gives practice; the student still has to build skill.
AI handles the draft; the worker owns the decision.
Alan Turing asks whether machines can think, turning AI into a serious question.
Deep Blue beats chess champion Garry Kasparov, showing machines can outperform humans in narrow symbolic games.
Deep learning breakthroughs make image and pattern recognition dramatically stronger.
ChatGPT brings generative AI into everyday conversation.
Claude and other systems become tools for writing, coding, research, schoolwork, and business pilots.
The question may shift from “Can AI do this task?” to “Who is responsible for the result?”
“AI doesn't remove teachers. It removes repetitive work.”
“Students still need someone who believes in them.”
“Automation saves time. Human connection creates learning.”
The thing that makes AI feel real in Seoul is that the city already runs on speed. Subway gates, delivery apps, kiosks, translation tools, and hagwon schedules all reward anything that saves time. That makes AI fit naturally into daily life.
But faster is not always better. A student who gets instant feedback can improve faster, but can also become dependent. A worker who uses AI to write can sound more polished, but maybe less personal. A teacher who uses AI to grade can save hours, but if the school only sees money saved, the human part of education can shrink.
My hope is that AI becomes like a really powerful pencil. It helps people think, draft, fix, translate, and practice. My fear is that companies use it as an excuse to remove beginners before they ever get the chance to learn.
That is why the future of work should not only be about which jobs disappear. It should be about which jobs become harder to enter, which skills become more valuable, and who gets access to the tools first.