<p>This article explores the design and implementation of AI-driven feedback systems to enhance student writing through a cultural-historical perspective. Drawing on Vygotsky’s and Galperin’s theories, the study emphasises the role of feedback as a cultural tool that mediates learning and fosters cognitive development. The article introduces the Essay Assessment Technology (EAT), an AI-powered tool developed to support formative assessment in writing. EAT provides structured feedback on language, content, and organisation, guiding students through iterative drafting processes. The study presents findings from three writing processes conducted in Norwegian secondary schools, comparing outcomes between classes using EAT and those relying on peer feedback. Results indicate significant improvements in student writing quality, particularly in classes utilising EAT, with large effect sizes observed in later writing processes. The article outlines seven pedagogical design principles for AI-driven feedback systems, emphasising the importance of aligning feedback with learning objectives, fostering student autonomy, and integrating social interactions. These principles aim to transform feedback from a static assessment tool into a dynamic resource that supports self-regulation and deeper engagement with the writing process. The study highlights the potential of AI to enhance formative assessment but underscores the need for balanced integration with human interaction to address higher-order writing skills and ethical concerns. By grounding AI feedback in cultural-historical theory, a framework is offered to develop technologies that not only improve writing proficiency but also prepare students for lifelong learning in an AI-driven world.</p>