In the era of accelerating technological change and systemic uncertainty, foresight has become a critical capability for organizations seeking to anticipate and navigate complex futures. At the same time, advances in generative artificial intelligence (GenAI) are rapidly transforming how foresight is conducted, enabling new forms of future exploration, data synthesis, scenario generation, and forward-looking analysis. Despite this growing relevance, the role of GenAI in foresight remains conceptually underdeveloped, with existing research predominantly framing artificial intelligence (AI) as a tool for augmenting human cognition. This paper addresses this gap by conceptualizing the role of GenAI in reshaping foresight practices and contributes to the development of futures literacy – the capacity to understand, imagine, and use the future in decision-making. Drawing on insights from futures studies, knowledge management, and socio-technical systems theory, this paper develops a conceptual framework that positions GenAI as a knowledge infrastructure mediating foresight processes. The paper’s main contribution is the identification of a dual-path transformation in AI-supported foresight. The first, the augmentation paradigm, conceptualizes GenAI as a cognitive collaborator that enhances human sensemaking, supports participatory foresight processes, and fosters futures literacy as a learning-oriented capability. The second, the autonomous foresight paradigm, reflects an emerging trajectory in which GenAI systems increasingly perform foresight tasks with limited human intervention, prioritizing efficiency, scalability, and accessibility. The paper highlights a fundamental tension between these two paradigms. While automation expands access to foresight and reduces costs, it may also weaken the reflexive and participatory processes that underpin futures literacy. This raises critical questions regarding the relationship between foresight production and anticipatory learning, as well as the role of human agency in AI-mediated knowledge creation. The paper contributes to the literature by advancing a novel conceptualization of AI-supported foresight as a socio-technical transformation characterized by competing epistemic logics. It further outlines implications for organizational practice, emphasizing the need for hybrid foresight models that integrate AI-driven capabilities with human-centered sensemaking, and identifies key directions for future research on governance, ethics, and human–AI co-creation in foresight.