"The threshold for entering the elite will become unrealistically high." Pavel Veinik on how AI will change career tracks

«IT is becoming a regular industry where people count money»

Two years ago, the career path for tech professionals seemed challenging but clear. It seems the context has changed since then?

Yes, indeed. Everyone’s anxious and trying to understand what’s next. There are many different predictions about AI that constantly change, and the situation is developing very, very rapidly. Anthropic has stated that, by the end of 2026, AI may be able to write complex systems on its own. We’ve become accustomed enough to miracles to consider this possibility.

This is unfolding against the backdrop of a global recession: money is no longer free, crazy venture investments remain mostly in AI, and the IT industry bubble that has been inflating since the late 90s is gradually deflating. It’s simply the cyclical development of the global economy.

IT is turning into a regular industry where people count money. Once, the IT boom in Belarus occurred due to the wage difference between a Belarusian and a European/American. Now this difference is smoothing out, and the call for «everyone to enter tech» is no longer relevant.

How should IT professionals develop in this situation? How relevant are your statements from two years ago about the two senior career tracks — technical and managerial?

Fundamentally, the division of roles in IT isn’t changing. The choice at the «tech or manager» fork is determined more by a person’s inclinations than by external circumstances. An introverted engineer (not to be confused with schizophrenia) often prefers the technical path simply because they enjoy technology. Management is preferred by extroverts. A technician can turn into a manager, but a manager is unlikely to become a technician.

AI is changing these roles, but not drastically. As engineers advance, their role requires more communication — not primarily managerial communication, but architectural communication: more coordination about system design, requirements and releases. They have to communicate more frequently about tasks, releases, etc. AI makes it faster to solve routine communication tasks. If company processes are well established, the additional communication load will be less noticeable. If processes aren’t structured, the increased number of communications can lead to chaos.

«An under-engineer with AI isn’t a career option»

Doesn’t AI open opportunities for yet another track — what’s called Solopreneurship?

Yes — an engineer with a solid technical foundation, a working understanding of AI, and some entrepreneurial ambition can automate many tasks that previously required an entire team. This is known as solopreneurship, which combines managerial, entrepreneurial and engineering skills. It’s easier to do now than before. If you’re a good engineer, of course. But still, this track is more about hype than reality for now.

You say an engineer with AI will survive, while one without AI probably won’t. Is there also a trend of «under-engineers» whom AI makes superficially productive at certain stages?

No, that’s not right. When it comes to simple things, we can take something like Lovable and create something that will somehow work. But if someone doesn’t have a background in distributed systems and understanding of architecture, and only knows how to create a form, a table, or a simple integration, they won’t be able to compete with an engineer. AI won’t help them manage complex things; the difference here is fundamental. So «an under-engineer with AI» is not a career option.

«Computer science is absolutely necessary»

AI seems to devalue academic foundations and technical knowledge. HR professionals mainly talk about communication, the ability to see business value, and using AI tools. But can one become a senior without fundamental knowledge?

I don’t observe devaluation of the technical foundation. On the contrary, I see that this bar is rising, fundamental knowledge is becoming even more important. Previously, a senior could write well in Java, use Spring and other frameworks — now that’s not enough to be a senior. AI can write code, but an engineer must tell it what code to write and for what purpose. For this, the engineer must mentally understand the logic of project development and the complexities that need to be considered. Often these are «minor details» that are easily lost due to the lost-in-the-middle phenomenon and other shortcomings of large language models.

Someone who doesn’t understand how a software project is created and developed will face enormous problems that will simply bury them. An engineer with deep knowledge and a broad perspective will be able to skillfully direct AI, in any language.

«It’s almost impossible for juniors to enter IT now»

Until now, one could enter tech by writing code and then master systemic knowledge. Now this track is under threat: nobody wants juniors — due to the crisis and AI. Especially when you can hire a mid-level engineer for the same money. Where will mid-level and senior engineers come from in a few years if the industry isn’t hiring juniors now?

Yes, seniors may start dying out after some time, creating specific risks. Now everyone has rushed to code their systems, and it’s not unlikely that only the old guard will remain who can maintain them. That battle-tested elite who are old-school but also know how to write prompts.

At the same time, the entry threshold to this elite will become incredibly high. It’s like becoming a lawyer in American films: first you’re an unpaid intern for 2-3 years, and only then do they start trusting you with something. But this is just my imagination; I won’t make predictions — it’s unclear how the industry, the economy, and the world in general will change.

It’s almost impossible for juniors to enter IT now; that time has passed. Until the industry stabilizes, there’s not much point in trying to enter it. For a junior to be minimally useful, they need to have a broad perspective and understanding of Computer Science.

A university graduate doesn’t have this?

No. Only in specific specialties, and even then only if they had the right professors. Computer science isn’t an immutable branch of mathematics; it’s a rapidly evolving discipline that university bureaucracies often struggle to keep up with.

But if university education doesn’t really help to enter IT, then what? Not IT courses, surely?

But if you’ve decided you want to enter IT not for the big money (which won’t be there anymore), but because your heart is in it, then you need to develop a learning program with the help of experienced IT professionals.

In my opinion, the main thing is systematizing your own thinking, a meta-skill that will help with everything. Such a program can be compiled from books on systems thinking and some courses. Yes, it’s difficult and without some external support, it’s practically impossible for a young person.

So I’ll be honest: I don’t know what juniors should do now.

«For AI to help business, you need clear processes»

IT professionals don’t like when employers force them to use AI. They say both code and communication suffer, while metrics grow exponentially along with stress. In general, a dilemma appears: quality and mindfulness or AI and speed.

AI disrupts the status quo for employees, and in this situation — naturally — people will complain. Some about being forced to use AI, others about being forbidden from using it. Such complaints are a reaction to anxiety, normal «sabotage.»

We don’t want to change, to master something new and unclear. This is our emotional response. But on the merits, when a company has clear processes, a sufficiently understandable work flow, it can be automated. When there’s some automation platform, even the simplest one, AI can be applied to it, which will help solve tasks systematically. Only under this condition will AI be useful to both the business and employees.

Setting up processes is difficult. In early 2025 a study reported that many large corporations invested heavily in systematically integrating AI into their processes. By year-end, only about 10% had succeeded; most had not.

So mindfulness is more about clear management processes. Then it won’t be opposed to quality.

«AI itself doesn’t provide professional growth»

Let’s summarize, does AI definitely provide professional growth and how not to confuse it with the race for KPIs?

The more creative the employee and the more intellectual their work, the worse KPIs work for them. Creative employees can hack KPIs easily. They can achieve metrics without essentially fulfilling the goals for which these metrics were created. A classic example is KPIs based on the number of lines of code.

And I don’t like the word «growth.» Artificial intelligence itself cannot provide growth; only a person can provide themselves with professional growth. What tools they use for this — another person, AI, a book, or something else — is a secondary question.

In this context, I like the words «effectiveness,» «productivity.» They can certainly increase.

Then what do you understand by professional growth?

For an engineer, it’s understanding distributed systems — everything that’s called Computer Science. The ability to hold systems of larger size in your head, taking into account a greater number of details.

And here AI tools won’t help at all — only the old traditional methods.

They won’t help. You either understand the system and build it, or you don’t build the system, and then you get a mess that falls apart under its own weight.

Can we then say that to the two vectors of IT engineer development — technical and managerial — AI simply adds an emphasis on productivity growth, but not professionalism?

Hmm. It seems AI confuses things here and makes us search for the right formulations again. Probably, you can’t increase productivity without enhancing professionalism. If I was an engineer without AI, and now I’m an engineer with AI and know how to use it, then this «know how» is the growth of my professionalism.

But this is not about Computer Science anymore.

OK, it’s not Computer Science, but something like literacy. Let’s say I’m a trench digger. I learned to read and now I can understand from paper where and how I need to dig. This increases my productivity in the system because less communication is required for me. It’s the same for an engineer who has mastered new skills in AI and can now perform their tasks better, faster, more efficiently. Even if it’s not Computer Science.

On the other hand, this thing greatly relaxes the brain and unlearns some basic skills.

For a person who tends to relax their mental activity, anything will relax it. For a person who tends to grow, anything will provoke mental activity.

For those who use AI to satisfy the simplest requests, it will be harmful. For those who build complex systems, it will bring benefits. This tool can be used for a huge number of tasks, including to become completely stupid. Everyone takes what they need.

Of course, this is a significant transition for civilization, and it will lead to social stratification. The world is moving in this direction.

Let’s discuss salaries in the context of these changes. During a crisis, you can hire people with mid-level experience for junior salaries and seniors for mid-level salaries. Does this mean that the glass ceiling for senior salaries is lowering?

When discussing salaries, geography, visas and other work-permit issues are critical. In Kazakhstan, for example, IT is growing rapidly, whereas in Belarus the market is weaker. That is, the recession story affects everyone, but differently in different countries.

Regarding this wage dumping. The engineer who, thanks to AI, has started coding less but solving more business problems remains very much in demand. Yes, it’s harder for them to break through various recruitment automations that use AI, but they’re still an important player. I haven’t heard about salary reductions for such people. I’ve heard about layoffs, but never about «now you’re doing the same as before, but getting paid less because the market has changed.» An experienced engineer remains in demand.

In the second echelon are those who do applied things like enterprise automation. They will earn substantially less, but still decently.

And then come all sorts of vibe coders, guys without a good background who automate various «bots» using no-code or low-code — this will be a long and cheap tail.

The complexity of the tasks that each of the three classes solves will gradually increase. And the engineer’s task is to end up in the highest class possible with the clearest and deepest knowledge. To remain in demand.