Industry

AI-generated text

AI could reshape tasks for most Hungarian workers, labour market map finds

A joint BeHive and AI Budapest project estimates that AI could significantly change the work of about 64% of Hungary’s employees over the next two to three years.

AI could reshape tasks for most Hungarian workers, labour market map finds

Around 64% of Hungarian workers hold jobs whose tasks could be substantially reshaped by artificial intelligence over the next two to three years. A labour-market impact map produced jointly by BeHive and AI Budapest suggests that this does not necessarily mean mass layoffs in the near term: in many cases, the mix of tasks and the skills required may change instead.

The creators examined 462 occupations and 4.67 million workers. The map is a Hungarian adaptation of Andrej Karpathy’s US model, drawing on data from the Hungarian Central Statistical Office (KSH), the FEOR-08 occupational classification and Cedefop. It assesses software-based AI, not the effects of physical robotics or industrial automation.

Hungary’s labour market has a different structure from those of the United States and Western Europe, so earlier estimates based primarily on those markets cannot be applied directly to Hungary. The Hungarian map divides occupations into three groups according to the type of change expected.

Three types of impact on jobs

About 414,000 people, or 9% of workers, are in the jobs expected to undergo the greatest transformation. They include administrative staff, accountants and entry-level software developers, whose work often involves repetitive tasks that follow defined rules.

In accounting, AI can process invoices, reconcile data and identify discrepancies. This does not necessarily eliminate the profession, but it may shift the emphasis towards more complex work requiring interpretation. Alongside learning to use the tools, workers may therefore need skills that go beyond tasks that can be automated. Companies also need to create the conditions for retraining.

The second group, described as augmentable, accounts for 54% of workers, or about 2.51 million people. It consists largely of skilled knowledge workers in fields such as law, finance and analysis. For them, AI is primarily a tool to complement and accelerate human work. A lawyer, for example, can use it to review lengthy documents, while professional judgment and the final decision remain human responsibilities.

Leaders in this group also need to decide which activities should be assigned to AI, where human involvement is essential and how workflows should be redesigned to improve productivity.

The third group, structurally stable jobs, includes around 1.75 million people, or 37% of workers. Their work is expected to change less as a result of software-based AI over the next two to three years. Many work in fields requiring physical presence or direct human contact, such as construction, cleaning and nursing.

That does not mean AI has no role in these sectors. At a construction company, for example, a bricklayer’s work may change relatively little while office administration becomes faster. Staff in these fields can therefore also benefit from preparation to use the technology as a support tool.

An open question for early-career training

Automation raises a particular question for people entering knowledge-work professions. Newcomers have traditionally gained experience through simpler tasks such as organising data or reviewing documents. Those are also among the activities AI can automate quickly.

It remains unclear how early-career workers will gain practical experience if, for example, a law firm assigns preliminary file reviews to a machine. Judgment and the ability to interpret context may become important earlier in their careers, but that could also bring greater responsibility than has previously been typical at that stage.

Technology alone does not guarantee productivity

Although AI can help many employees work faster, corporate investments have in many cases yet to deliver the expected measurable productivity gains. Building and operating infrastructure is costly, and revenue does not always keep pace with spending. In a BCG study covering 59 countries, around three-quarters of executives surveyed said they struggled to demonstrate tangible business value from using AI.

The difficulty is not necessarily the technology itself: even successful pilot projects can stall if workflows and decision-making do not change alongside them. The source identifies three recurring organisational obstacles:

  • An expectations gap: leadership strategy is not translated into the daily tasks of individual roles.
  • A trust deficit: employees are uncertain how to check AI outputs and who is accountable for them.
  • A measurement gap: without feedback, companies cannot see where adoption is stalling or whether it is creating business value.

Anthropic data also shows a gap between AI’s theoretical capabilities and its actual use. Large language models could theoretically be applied to 94% of tasks in US business and finance occupations, while actual Claude use is observed in 30% of those tasks. The figures concern the US labour market and use of Claude, so they cannot be applied directly to Hungary.

Preparing for different kinds of change

Over the next two to three years, companies will need to address jobs with tasks that can be automated, jobs that AI can support and jobs that are changing less for now. In the short term, that means reconsidering existing tasks; over the longer term, it may also involve testing workflows built around more autonomous, agentic AI systems.

This requires clarity about what AI can do and where human decisions remain indispensable. Alongside safe technical conditions, training needs to cover more than basic tool use: it should also help employees understand how their own roles are changing. Ultimately, productivity gains will also depend on whether organisations can bring technology and human expertise together in everyday work.