AI adoption doubles in EU, but only few countries are driving the growth

A new data-driven analysis has exposed growing disparities in artificial intelligence (AI) adoption across European Union member states, revealing a continent divided between digital frontrunners and lagging economies. Published in Economies under the title "Artificial Intelligence Adoption in the European Union: A Data-Driven Cluster Analysis (2021–2024)", the study provides the most comprehensive enterprise-level investigation to date into how AI technologies are spreading, which barriers persist, and how national strategies are shaping outcomes.

Drawing on Eurostat’s ICT usage surveys from 2021 to 2024, the research uses principal component analysis (PCA) and K-means clustering to quantify adoption typologies, identify technological specialization, and track divergence among EU states. While AI adoption by enterprises nearly doubled in this period, rising from 7% to 13%, progress remains uneven, with many countries struggling to keep pace amid skills shortages, rising costs, and regulatory uncertainty.

How has AI adoption progressed across the EU and which countries are leading?

Despite an overall increase in AI adoption, the EU-wide average remains low when compared globally, particularly against the United States and China. Denmark consistently emerged as the EU leader in enterprise AI uptake, with 2024 adoption rates reaching 27.6%, followed by Belgium, Sweden, and Finland. In contrast, Romania, Poland, and Bulgaria posted persistently low rates, all below 7%.

Early in the observed period, enterprises gravitated towards process automation, machine learning, and text mining. However, the debut of generative models such as ChatGPT in 2022 triggered a surge in interest in natural language generation (NLG), which became the most adopted AI subtype by 2024. Other growing categories include speech recognition and workflow automation, particularly among digitally mature economies.

Cluster analysis over time revealed a narrowing divide in strategic AI priorities. By 2024, three dominant adoption clusters had emerged: countries with high, diversified AI use; those specialized in narrow applications such as image recognition; and those with systematically low adoption. Denmark, Belgium, and the Netherlands formed the core of the high-adoption cluster, while countries like Malta and Slovenia displayed strong specialization in limited AI domains.

What are the key barriers preventing broader AI integration?

The study highlights that adoption barriers have shifted over time but remain stubbornly entrenched. The most consistently cited hurdle across all years was a lack of relevant expertise, reported by 7.1% of enterprises in 2024. Cost-related challenges climbed sharply in significance, particularly among small and medium-sized enterprises. Other persistent obstacles included poor data quality, legal ambiguity, and concerns over privacy and data protection.

Regulatory clarity also plays a pivotal role. In the wake of the EU’s AI Act, enterprises continue to struggle with implementation uncertainty. While some countries, particularly those in the high-adoption cluster, have begun to develop internal capabilities, a majority still rely heavily on externally developed AI solutions. In 2024, 84% of AI-using enterprises depended on commercial or open-source tools, underscoring a deep dependency that may expose the region to external strategic vulnerabilities.

Bias management in AI tools remains inconsistent. Enterprises in Belgium, Denmark, and Malta showed greater attention to bias checking, especially for AI models modified internally or sourced from open platforms. However, many others reported lower vigilance, particularly where AI was deployed in ready-made commercial form.

How should EU policymakers respond to a fragmenting AI landscape?

The study concludes with a series of policy recommendations based on the identified clusters of AI adoption. For digital leaders like Denmark and Finland, the focus should now shift to enhancing R&D, supporting small business adoption, and refining ethical frameworks. Countries in this group are urged to create regulatory sandboxes and data-sharing platforms that reinforce trust and facilitate innovation.

Lagging countries such as Romania, Bulgaria, and Poland, on the other hand, require foundational investments in digital infrastructure, specialized workforce training, and access to AI development tools. Tailored national AI strategies with clearly defined goals and measurable milestones are essential. The authors stress the importance of using EU funding instruments to bridge the AI gap through incentives, pilot projects, and public sector deployments that demonstrate tangible benefits.

Meanwhile, niche-focused adopters like Malta and Slovenia are encouraged to leverage their specialization by creating centers of excellence in their dominant AI verticals. These countries should also broaden their application scope through targeted education, regulatory flexibility, and cross-border partnerships.

Importantly, the study warns against a one-size-fits-all policy model. The depth of divergence in adoption patterns, both in terms of technology and purpose, demands differentiated strategies. Uniform mandates risk exacerbating inequalities rather than reducing them.