DIGITAL CAPABILITY AND AI ADOPTION IN MEDIUM-SIZED ENTERPRISES IN THE EU

Abstract

Adoption of AI varies across the EU, even though the differences in ICT training and employment of ICT specialists are smaller. This shows that the simple presence of digital resources does not ensure the effective integration of the artificial intelligence into enterprises' processes. The study analyses the relation between digital capability and AI adoption in medium-sized enterprises in the EU. Digital capability is evaluated through two elements: employee training in ICT and the employment of ICT specialists. Romania's position is also analysed separately. The analysis uses Eurostat data for 2024 on enterprises with 50–249 employees. The main sample included 26 EU Member States. Descriptive statistics, Pearson correlations and five linear regression models were used. Diagnostic and sensitivity checks were also performed, and countries were grouped descriptively according to their digital profile. Romania’s position was evaluated by comparing it with European averages, indicator rankings and the values ​​estimated by the main model. Both ICT training and employment of ICT specialists are positively correlated with AI adoption. The connection is stronger for employee training, with a correlation coefficient of r =0.708, compared to r =0.642 for ICT specialists. When the two variables are analysed together, training remains statistically significant (B=0.418, p=0.023), while ICT specialists are no longer significant (B=0.273, p=0.225). The sample explains 53.3% of differences between countries. Romania has an AI adoption rate of 3.88%, the lowest in the sample The results show that the development of digital skills among employees is more closely connected to the adoption of AI than the simple employment of ICT specialists. Technical expertise remains important but is more effective when supported by constant employee training. Because the analysis uses aggregated data for a single year, the results indicate associations across countries, but do not demonstrate causal relationships.