"The phrase 'live forever, learn forever' has never been as relevant to the IT industry as it is now." What will remain valuable for a programmer when the code is generated by a machine

«It’s not the skill that depreciates, it’s the approach»

Most of dev.ua’s interlocutors categorically disagree with the statement «AI is taking jobs away from IT people.» In their opinion, the problem now is not that AI is displacing people, but that the market is ceasing to value purely mechanical code writing.

Timur Bolotyukh from PlantIn puts it even more harshly: «AI devalues ​​not the skills of engineers, but the performance of routine work. What is devalued is the approach in which the engineer only knows how to write code, but does not solve the problem comprehensively.»

And this is perhaps the main shift that is currently taking place in the market. If earlier companies could hire a strong «task performer», now they are increasingly looking for people who understand the context, see the system as a whole and can make decisions. Olena Oleksienko, Chief People Team at EVO, speaks of this as an established market trend: demand is shifting from simple delivery to roles that affect business results.

Code generates LLM. Architecture is still people

Interestingly, almost all the interlocutors separately mention systems and architectural thinking. It is now called the main skill that AI cannot yet replace. The reason is quite simple: an LLM can generate a piece of code, but it does not think about what will happen to the system in a year, when the load will triple. AI, the interlocutors of the publication say, does not know the business limitations of a particular product and does not understand the trade-offs between speed, scalability, price and security.

«AI generates solutions, but does not know the context, limitations, or features that the client may have regarding system load. The task of a person with architectural thinking is to think ahead, weigh realities, and take into account various factors,» explains Vlasenko.

PlantIn adds: the greatest value of an engineer now is in the ability to build reliable systems on top of fundamentally unreliable components, because generative AI still often makes mistakes, hallucinates, or offers solutions that look nice in a demo but don’t work well in real production.

AI is good at writing code. But it’s bad at understanding the task

Another thesis from the interviewed experts is that the main problem of most AI experiments is not even the quality of the code, but that people poorly formulate the tasks. Many teams have already encountered a situation where the developer simply «dumps» the entire context into ChatGPT or Cursor and expects a magical result. But AI works much better when the task is properly decomposed: there are clear boundaries, interfaces, quality criteria and understandable restrictions.

That is, an engineer who can decompose a chaotic business requirement into a clear system of subtasks will work with AI much more effectively than a person who simply copies prompts from Reddit.

T-shape instead of narrow specialization

Another thing that the interlocutors almost do not argue about: narrow specialization is becoming a less safe strategy. The market is moving towards T-shaped specialists — people with strong core expertise, but with an understanding of related areas. And here, on the contrary, AI can be a huge accelerator.

A similar opinion is expressed by Igor Kozlov, Data & ML Engineer at Levi9: «Even with the use of AI agents, there should still be a deep understanding of what exactly you are building. This allows you to see risks, weaknesses, or stop the agent in time if it has gone the wrong way.»

AI fluency becomes the new basic literacy

A few years ago, resumes used to say «confident PC user.» Now the market is gradually moving towards a new minimum — the ability to work normally with AI tools.

Daryna Kuzmyk, Talent & Culture Lead at Railsware, ironically says: «Just as it was once customary to write ‘confident PC user’ on a resume — although often this confidence was hidden only by the ability to find the ‘Start’ button — so now we need ‘confident AI users’, for whom AI is not just ‘Google at maximum speed’.» But the ability to write a prompt has long ceased to be a competitive advantage. It is basic hygiene. The real difference is in the ability to evaluate the result, understand where AI is wrong, check the quality, and take responsibility for the final decision.

EPAM calls such people «qualified AI customers.» That is, those who don’t just press Enter, but understand the limitations of the model, identify checkpoints, and see risks before they even make it into production.

And this is where another critical skill comes in: learning speed. Today everyone is talking about LangGraph. Tomorrow the market will switch to something else. The winners will not be those who have learned one specific tool, but those who are able to adapt quickly.

Security awareness is a problem that many underestimate

A separate topic that is being actively discussed by IT company representatives is the security of working with AI. And while most teams are enthusiastically testing agents and generating code, some companies are already thinking about the risks of data leakage, access, and internal information.

Levi9 already sees a demand for a separate specialization — AI Security Engineer, because the problem is no longer just using AI, but understanding when to stop it.

Business is no longer just about «the person who writes code»

Almost all the companies that dev.ua spoke with describe the same change: engineers are increasingly being evaluated not by the amount of code written, but by their impact on the business. At Ciklum, for example, they directly talk about the movement from a «developer» to a person who manages complex digital systems and AI tools. At EVO, they add: if in 2021 the market «hired potential,» now it hires results. Therefore, real cases, pet projects, open source, or even strong volunteering often weigh more than a beautiful list of certificates.

And at the same time, according to hiring experts, the value of what many called «soft skills» a few years ago is growing: communication, problem-solving, analytical thinking, adaptability, because technologies are changing faster than any stack can become «stable.»

And it seems that the main change in the market now is not even that the machine has started generating code, but that the ability to write code itself is no longer enough. The engineering base is not disappearing anywhere, but on top of it, systems thinking, AI fluency, business understanding and the ability to adapt quickly are now almost mandatory. This is what seems to be the new «seigniorage» of the coming years.

Senior inflation in IT: why years of experience no longer sell specialists to employers, how companies determine candidates' seniority and who has a better chance of getting an offer

On the topic

Senior inflation in IT: why years of experience no longer sell specialists to employers, how companies determine candidates' seniority and who has a better chance of getting an offer

Read the country’s main IT news in our Telegram

On the topic

Read the country’s main IT news in our Telegram