The fluorescent lights hummed in the small office building near Belgrade's Autokomanda junction. Inside, a young man named Marko, barely out of university, meticulously drew bounding boxes around images of traffic signs. His task was to identify every stop sign, every pedestrian crossing, every speed limit marker in a seemingly endless stream of photographs. He was not building an AI, not coding an algorithm, but he was feeding the beast. This is the often-unseen reality of the AI boom, a global industry built on the painstaking, repetitive work of data labeling, and Belgrade is playing a surprisingly significant role.
For years, the narrative around AI has been dominated by grand pronouncements about artificial general intelligence, self-driving cars, and chatbots that write poetry. But beneath the glittering surface of innovation lies a vast, labor-intensive infrastructure: the data labeling industry. Companies like Scale AI, Appen, and Sama have built empires by orchestrating human annotators to clean, categorize, and label the colossal datasets required to train today's sophisticated machine learning models. Without these labeled datasets, the most advanced algorithms are just empty shells.
Let's talk about what's actually working. According to a recent report by Grand View Research, the global data annotation market size was valued at approximately $1.6 billion in 2023 and is projected to grow at a compound annual growth rate of over 25% through 2030. This isn't theoretical growth; it's tangible, job-creating growth. While much of this work has historically been outsourced to countries with lower labor costs in Asia and Africa, the demand for higher quality, more nuanced, and often language-specific labeling has brought a significant portion of it to Eastern Europe, including Serbia.
The Balkans have a different relationship with technology. We've seen trends come and go, from the dot-com bubble to the crypto craze. What sticks are things that create real jobs and integrate us into the global economy in a meaningful way. Serbia's tech sector, traditionally strong in outsourcing software development, has found a new niche. Companies here are not just consuming AI, they are actively contributing to its creation, albeit at a foundational level. This is not about flashy startups with billion-dollar valuations, but about steady, reliable work for a growing workforce.
Data from the Serbian Chamber of Commerce indicates that the IT sector continued its robust growth in 2025, with exports exceeding €3.5 billion, a significant portion of which is attributed to services including data processing and annotation. While specific figures for AI data labeling are hard to isolate, industry insiders confirm a substantial uptick. Marko, for instance, works for a local firm that partners directly with a major global data labeling provider, supplying annotated data for autonomous vehicle development for a well-known German automotive giant.
Who are the winners and losers in this quiet shift? The clear winners are the local outsourcing firms that have been agile enough to pivot into data labeling. They are securing contracts, expanding their teams, and offering entry-level tech jobs that did not exist a decade ago. For example, companies like DataOps and LabelSrbija, though smaller players on the global stage, are growing their employee count steadily, providing stable employment in a region where economic stability is highly valued. The losers, if you can call them that, are perhaps those who remain fixated solely on high-end software development, missing the opportunity in this burgeoning foundational layer of AI.
From a worker's perspective, the reactions are mixed but generally positive. For many, it represents a foot in the door to the tech industry.
