AI Is Changing Entry-Level Technology Careers
By Suad Seferi ·
For years, young people across the Balkans were encouraged to learn programming. Technology offered something increasingly difficult to find in the region: competitive salaries, international clients and the possibility of building a career without necessarily leaving home. Coding academies expanded, computer science programmes attracted more students, and software development became one of the clearest routes towards economic security. That promise has not disappeared. But the path towards it is becoming less certain. A new generation of AI coding tools can now inspect software projects, write functions, fix errors, create documentation and complete tasks across multiple files. They are no longer limited to suggesting the next line of code. Increasingly, they can receive a description of a problem, develop a plan and carry out much of the implementation. For experienced engineers, these systems can save time and remove repetitive work. For companies, they can make smaller teams more productive. For graduates trying to secure their first position, however, the change raises a more difficult question: what happens when some of the work traditionally given to junior employees can be delegated to an AI agent? The first rung of the career ladder Entry-level work has never been only about productivity. Junior developers are often assigned smaller bugs, routine testing, documentation and relatively simple features. These tasks may not be the most complex parts of software development, but they allow new employees to understand a codebase, learn from mistakes and gradually take on greater responsibility. They are how many senior engineers became senior engineers. If companies automate too much of that work, they may reduce costs in the short term while weakening the process through which the next generation develops professional experience. Reporting from Silicon Valley already shows growing anxiety among graduates attempting to enter the technology sector. Young candidates with strong academic backgrounds are facing a market shaped by cautious hiring, corporate restructuring and intense competition for specialised AI talent. At the same time, companies are asking what work can be completed by smaller teams supported by increasingly capable tools. The evidence does not support the simple claim that AI has already destroyed the technology job market. Economic uncertainty, overhiring during earlier technology booms and changing investment priorities are also affecting recruitment. But AI is becoming part of the calculation. Research examining AI-native startups found that these companies tended to operate with smaller teams and employ fewer entry-level workers and managers than comparable businesses. They continued to hire engineers, but concentrated more heavily on highly skilled and specialised candidates. This points towards a labour market that may not contain fewer opportunities in every area, but could become more difficult to enter. Coding is not disappearing The arrival of coding agents does not mean software engineers are becoming unnecessary. Current systems can produce convincing code that contains security weaknesses, misunderstood requirements or problems that only appear when the software is deployed in a real environment. Someone still needs to define the problem, understand the wider system and decide whether the proposed solution is reliable. Even Anthropic’s own research on agentic coding presents a more measured picture than the idea of fully autonomous software development. Developers reported using AI throughout a large part of their work, but said that only a relatively small share of tasks could be delegated completely. Human supervision, validation and technical judgement remained necessary. The developer’s role is therefore not simply disappearing. It is changing. Engineers are increasingly being asked to explain requirements clearly, divide complex problems into manageable tasks, supervise AI-generated work and recognise when an apparently functional solution is wrong. As agents take on more implementation work, verification may become one of the most important software engineering skills. Researchers studying the future of the profession have similarly argued that testing, validation and the ability to assess AI-generated output will become increasingly important. This creates an uncomfortable situation for new developers. The people best equipped to supervise an AI coding system are often those who already understand software architecture, debugging and the consequences of poor technical decisions. Those skills are normally developed through years of practical work. A beginner may be able to generate software more quickly than before, but speed is not the same as understanding. A different challenge for the Balkans The consequences could be particularly important for the Balkans, where technology has become one of the region’s strongest connections to the global economy. Companies across North Macedonia, Serbia, Albania, Kosovo, Bosnia and Herzegovina, Bulgaria, Romania, Croatia and neighbouring markets provide software development and technical services to international clients. The sector has created opportunities for young people who might otherwise have looked for work abroad. Part of the region’s growth has been built on access to capable engineers at costs below those in Western Europe or the United States. AI changes that equation. When a small team in London, Berlin or San Francisco can use coding agents to complete more work, outsourcing decisions may no longer depend only on the price of developer hours. Balkan companies will need to compete through specialist knowledge, reliability, communication and a deeper understanding of the industries they serve. This does not make the region irrelevant. Balkan engineers have repeatedly demonstrated that they can build products for international markets and adapt quickly to new technologies. But adaptation cannot mean…