This study introduces a data-driven approach to measure and enhance digital competencies in Online Collaborative Learning (OCL) environments. Recognizing the critical role of digital competence for student engagement and collaborative success, this research leverages Learning Analytics (LA) and Large Language Models (LLMs) within a Conversational Agent (CA), “EduBot.” EduBot analyzes user activity data and provides tailored feedback to triger self-reflection for competence development. Experts identified 57 behavioral patterns of digital competence, of which five were operationalized based on alignment with available data types and measurability. Integrated into an OCL framework, EduBot collected and analyzed activity data from student interactions, providing feedback on communication, ethical practices, and collaborative tool use. The evaluation involved 20 students, revealing mixed responses: some participants found EduBot beneficial for self-reflection, while others raised concerns about feedback accuracy. Key findings highlight EduBot's potential in fostering digital competencies, though challenges in data precision and CA personalization remain. This research contributes a foundational model for competence-based feedback in digital learning and emphasizes the need for refined data collection to fully capture complex digital behaviors.