The formal is inextricably linked to the duality of the informal, which derives both from the inherent characteristics of an individual and from the interaction with the external environment. Informal employment is not homogeneous and shows a distinct segmentation. This paper investigates potential of artificial intelligence (AI) algorithms to profile and predict informality under conditions of limited data availability. Using Eurobarometer 92.1 survey data (2019) on undeclared work in the European Union, enriched with macroeconomic indicators, the study analyzes the profiles of individuals working without contracts, receiving envelope wages, or engaging in false self-employment. By comparing AI-based approaches (MLP, CNN, TabNet) with the machine learning benchmark Random Forest, we show that advanced AI models deliver competitive predictive performance but do not substantially outperform established algorithms for tabular data. The contribution lies not in claiming causality, but in demonstrating how computational models can uncover patterns of association between socio-economic factors and informal behaviors, thereby complementing traditional economic approaches. The findings highlight the role of poverty, education, occupational structure, and institutional trust in shaping informal labor, offering policy-relevant insights while recognizing the predictive, not causal, nature of the models.