"Rapidly emerging labour-market trends, new ways of working, and careers that build on continuous learning demand a new generation of skills intelligence." These words from Cedefop Executive Director Jürgen Siebel capture both the urgency and the ambition driving a fast-moving field, one where the raw material of job vacancy text is being transformed into some of the sharpest real-time insights available on European labour markets.
A decade of digging into vacancy data
Online job advertisements (OJAs) have long been a cornerstone of Cedefop's skills intelligence work. For over a decade, the agency has mined vacancy data to track how digital and green transitions translate into concrete skill demands, to identify emerging competencies before they surface in formal taxonomies, and to understand labour market dynamics with more granularity, supplementing traditional surveys. That work, developed jointly with Eurostat through the Web Intelligence Hub, underpins tools like Skills-OVATE, which provides continuous, cross-country insights on occupational skill needs.
A field at an inflection point
But the field is at an inflection point. Researchers from public institutions, international organisations, universities, research centres and the private sector gathered recently in Thessaloniki to take stock and to look ahead. The breadth of contributions reflected how far the discipline has matured. Work ranged from studies on the twin transition and AI's impact on the labour market, to personalised career-pathway recommendations built on OJA analysis, to investigations of gender bias encoded in vacancy language. Methodological advances in natural language processing are opening new possibilities for extracting skills signals, building taxonomies, and even analysing how individual competency profiles evolve over time. Regional analyses are bringing skills intelligence closer to the local realities of workers and employers. And a growing body of work is examining how OJAs can serve as early indicators of structural labour-market change — including the shocks of COVID-19, the war in Ukraine, and the accelerating twin transition.
A human sovereignty challenge
Yet the most provocative question on the table was not technical. Michail Skaliotis, Active Senior at the European Commission, put it starkly: "The next decade of OJA research is not a data science challenge; it is a human sovereignty challenge." As AI increasingly mediates the relationship between workers and employers — matching competencies to opportunities, generating job descriptions, filtering candidates — the nature of what an online job advertisement even is, is changing. It is becoming less a static vacancy notice and more a dynamic node in an automated ecosystem. That shift has profound implications for research, for policy, and for people. Under the EU AI Act, AI systems used in recruitment and employment are classified as high-risk, meaning transparency, human oversight and non-discrimination are legal requirements, not aspirations. In this context, OJA research is no longer only about understanding labour market demand, it is also about generating the evidence needed to scrutinise how AI-mediated labour markets function. The research community is, in effect, helping to build the evidence base that makes such oversight possible.
Questions about algorithmic bias in AI-generated job descriptions, the rise of skills-first hiring architectures, the spread of labour-market activity to channels that leave no public data trace — these are not edge cases. They are the defining challenges of the next phase. And they point to something that no model can resolve alone: the need for transparency, interoperable data infrastructures, and governance frameworks that keep human interests at the centre.
The data is richer than ever. The harder work is deciding what to do with it.