This article examines the role of artificial intelligence in academic assessment, with particular focus on its implications for grading accuracy. It presents a comparative analysis of 46 undergraduate reports, each evaluated independently by a human instructor and an AI-based system. A quantitative methodology combining correlation analysis, Bland-Altman plots, and paired t-tests was employed to compare human and AI evaluations.The results reveal discrepancies between the two sets of scores, which suggest that AI systems fail to adequately assess qualitative dimensions such as originality, critical thinking, and contextual relevance. While AI offers advantages in consistency, scalability, and efficiency, it often fails to capture the pedagogical depth and interpretive insight provided by human evaluators. In light of these findings, the article proposes a hybrid assessment framework that combines the precision of AI with the interpretive competences of human instructors.