This paper examines the use of Duolingo, an artificial intelligence-based language learning tool, to support teachers in managing large, heterogeneous classes through differentiated instruction. Unlike most studies that focus primarily on student progress, this research centers on the tool’s impact on teaching practices, specifically in tracking student needs and managing diverse proficiency levels. Conducted in a rural classroom of 40 A1-A2 level students, 20 volunteers engaged in weekly Duolingo sessions over three months, under close observation. Employing a qualitative approach with participant observation and a summative test, the study captures how Duolingo’s adaptive learning features enabled the teacher to monitor individual progress, tailor instruction, and efficiently address skill gaps—a challenging task in traditional classroom settings. Findings reveal that 85% of students maintained consistent engagement throughout the study, with four out of six initially low-performing students reaching an intermediate level by the final assessment, and two of them ending in the podium of earned xp. The results highlight Duolingo’s potential as a valuable complement to traditional teaching methods, enabling teachers to offer individualized learning pathways in large-group settings. Several limitations emerged, including the limited sample size, and the restricted capabilities of Duolingo itself. These findings underscore the need for an AI tool explicitly designed to support adaptive learning in classroom contexts, tailored to the specific needs of educators.