Artificial intelligence (AI) is increasingly positioned within mathematics education discourse as a potential resource for supporting learner engagement and conceptual exploration across core mathematics topics. In parallel, constructivist pedagogy continues to emphasise learning as an active, reflective, and socially mediated process grounded in learners’ meaning making and interaction. Despite these parallel developments, existing conceptual scholarship rarely integrates AI and constructivist mathematics pedagogy in ways that attend explicitly to curriculum content, educational phase, and contextual implementation conditions. This limitation is particularly pronounced in Global South contexts. This article is explicitly conceptual in nature and addresses this gap by advancing an AI-enhanced constructivist mathematics pedagogy framework focused on core mathematics topics in the Further Education and Training (FET) phase in Gauteng North. The study draws on a systematic theoretical synthesis of literature spanning constructivist learning theory, digital pedagogy, and implementation science. Within the framework, AI is theorised as a pedagogical mediator rather than a driver of instruction, with attention given to how adaptive tools can be aligned with individual meaning making, how collaborative platforms can support dialogic knowledge construction, and how implementation science foregrounds issues of adoption, sustainability in under-resourced educational settings. The framework constitutes the central scholarly contribution of the article, integrating pedagogical, curricular, and systemic considerations while foregrounding teacher agency, ethical responsibility, and contextual readiness. Although grounded in the FET mathematics context of Gauteng North, the framework is conceptually transferable to comparable educational systems and provides a basis for future empirical inquiry, theoretical refinement, and context-responsive pedagogical design in mathematics education.