AI and work: what has already changed and what to teach your child
AI has not caused mass unemployment, but fewer young people are hired. The data, the situation in Uzbekistan and practical steps for parents.
Ruslan Yuldashev, co-founder of Mars IT School
Worked in Yandex's education department in Moscow and at Silicon Valley startups Mavrx and GetGoing (San Francisco)
In short (data as of October 13, 2026):
- AI has not caused mass unemployment so far, but companies hire fewer young people: in the occupations where AI takes over the most tasks, employment among 22-25-year-olds is 19% lower than it would be had it grown in line with less-affected occupations;
- nobody knows how fast this will go further: forecasts range from "there will be more jobs" to "10-20% unemployment within five years";
- skills that AI cannot replace are rising in value: finishing the job, setting a task clearly, checking the result.
I run several companies, and one of them is the Mars programming school. So I look at AI from two sides: as an employer and as someone responsible for what children learn. Below is what I see myself and what the data shows.
What I see in my companies
AI took on programming first, but it will not stop there: next the same story moves into analytics, marketing and finance.
What the data shows
My experience matches the research, with an important correction: the blow fell not on everyone but on those just entering a profession.
- Fewer young people are hired. Stanford, using US payroll data: in the occupations most affected by AI, employment among 22-25-year-olds is 19% lower than it would be if it had moved in line with the rest. Experienced workers show no such gap. Companies are not laying people off, they are simply taking on fewer newcomers.
- There is no mass unemployment. Anthropic and Yale University found no rise in it even in the most affected occupations.
- Programmers are affected more than anyone. According to Anthropic, AI already performs about 75% of a programmer's typical tasks. But inside the profession the picture varies. Here is the US official statistics forecast to 2035:
| Role | What they do | Forecast to 2035 |
|---|---|---|
| Programmer working to a brief | writes code from a ready specification | −7% |
| Software developer | understands the problem, designs, checks | +10% |
The work of "write code from a ready brief" is shrinking. The work of "understand what is needed, design it and check it" is growing.
What about Uzbekistan
The state is betting on AI seriously. A strategy through 2030 was approved in 2024. In summer 2026 a supercomputer built on NVIDIA technology started working in Tashkent. In the global ranking of government readiness for AI, the country rose from 87th place to 62nd in two years.
The money in the profession is real too. Average salary according to the National Statistics Committee for the first quarter of 2026:
| Sector | Average monthly salary |
|---|---|
| Programming and consulting | UZS 21.2 million |
| Tashkent, all sectors | UZS 11.7 million |
| Country average | UZS 6.8 million |
But schools do not yet teach AI systematically. One pilot is confirmed: School No. 9 in Yunusabad district, grades 9-11, since June 2026. For now this is left to parents.
What comes next: honestly, nobody knows
The World Economic Forum expects 92 million jobs to disappear by 2030 and 170 million to appear. Anthropic's head Dario Amodei warns of 10-20% unemployment within 1-5 years. Economists mostly speak of changes over ten years or more. The argument is mainly about speed. The direction is clear to everyone.
When something becomes cheaper and more accessible, people consume more of it, not less. Creating software has become cheaper, so there will be more software.
This is how I answer parents. Programmers have always studied all their lives, because technology kept changing. Now it changes faster. But there is an old observation by economists, the Jevons paradox: when something becomes cheaper and more accessible, people consume more of it, not less. That is what happened with coal in the 19th century. Creating software has now become much cheaper, and demand for business automation is growing. More people who can do this professionally will be needed, but different people than before. This is my estimate, not a proven fact.
Which skills are rising in value
- Finishing the job. Starting with AI is easy. Building a working thing, finding the errors and delivering it is a rare skill.
- Setting a task clearly. Someone who can state precisely what is needed gets many times better results from AI. This is a thinking skill, not knowing which buttons to press.
- Thinking logically and checking. Only 48% of developers trust an AI answer, and only when they can verify it (Stack Overflow 2026 survey). Only someone who understands the subject can verify.
- Managing agents. AI agents are becoming doers. People are needed who give them tasks and answer for the result.
The World Economic Forum names analytical thinking as the top skill: 7 out of 10 companies consider it essential.
Careful: AI can also harm learning
In Turkey an experiment was run on about a thousand high school students. Those who solved math problems with ordinary ChatGPT then wrote an exam without AI 17% worse than students who studied without it. Those given an AI tutor that does not hand out ready answers suffered no harm.
The conclusion is simple: what matters is not whether AI is present but how it is used. If AI solves things for the child, the child does not learn. That is why the rule at Mars is: the child is the author, AI is a tool, and the child must be able to explain every line of their project.
What a parent can do right now
- Do not ban AI and do not hand it the homework. Suggest your child make something for themselves with AI.
- Start with a project the child actually needs. In the first month at Mars, in "AI-Pilot", children make things like: a practice tool for English, Korean or Arabic, a tracker for useful habits, videos and presentations.
- Ask them to explain: how it works, what you asked AI for, where it made a mistake. If a child can explain, they are learning.
- Praise finishing, not starting.
If you want this done systematically and with a mentor, come to a trial lesson: marsit.uz/en/courses.