This study developed and evaluated a design framework that embeds Artificial Intelligence-Driven Personalized Learning (AI-DPL) in the e-IM3 learning model to strengthen the numeracy of elementary school students. Design-Based Research (DBR) was the guiding framework, and an exploratory sequential mixed-methods strategy organized data collection and analysis in three iterative phases: analysis and exploration, design and construction, and evaluation and reflection. The research took place in an elementary school in Probolinggo Regency, East Java, Indonesia, with 10 fourth graders in a limited trial and 20 fifth graders in a field trial. Observations, semi-structured interviews, focus group discussions, and document analysis provided the qualitative data, whereas expert validation, practicality questionnaires, and numeracy pretests and posttests provided the quantitative data. Three expert validators judged the framework highly feasible, and its content validity was high (S-CVI/Ave = 0.96). Numeracy scores rose significantly in the limited trial (N-Gain = 0.733) and in the field trial (N-Gain = 0.747; Wilcoxon p < 0.05), and practicality reached 93.6%. Since a one-group pretest–posttest design was employed, these outcomes point to improvement, not to causal effectiveness. Overall, combining AI-driven personalization with the e-IM3 pedagogical syntax appears to be a promising route for numeracy learning through adaptive instruction, AI-assisted scaffolding, and learning analytics, and the study offers refined design principles for learner-centered numeracy instruction in elementary schools.