Digital transformation has introduced technologies such as artificial intelligence (AI), automation, and learning analytics into education, generating more personalized, efficient, and data-driven learning environments. These innovations support intelligent tutoring systems, automated feedback, performance prediction, and administrative decision making. However, their implementation also presents ethical, social, and technical risks that remain insufficiently understood and regulated. This study presents a narrative review of scientific literature published between 2021 and 2025, aiming to analyze the uses, benefits, and challenges of AI in educational contexts, with a particular focus on learning analytics. Articles were selected from major databases such as Scopus, IEEE, SpringerLink, and Web of Science, prioritizing studies that offered practical applications and a critical lens. The review highlights promising developments, including systems that enhance personalized learning trajectories, reduce grading time significantly, and predict student dropout risk with high accuracy. Nonetheless, it also exposes significant concerns, such as the reinforcement of algorithmic biases, excessive surveillance, and the exacerbation of digital inequalities—especially in Latin America, where a substantial portion of students still lack stable access to quality Internet. These findings underscore the dual nature of AI in education: while it has the potential to improve quality and equity, its effectiveness depends on ethical governance, equitable infrastructure, and meaningful teacher training. The review concludes that AI should be used to complement, rather than replace educators’ roles, placing emphasis on learner needs, human interaction, and contextualized pedagogy. Achieving a more human-centered and inclusive education will require a deliberate balance between technological innovation and professional educational judgment.