The pressing development of generative AI has prompted a surge in future-oriented documents anticipating extensive educational impact within the Nordic welfare state. Moulded in an anticipatory regime of future thinking and living (Adams et al., 2009), these documents shape what futures ought to be and cast them as inevitable, making it a key concern for education scholarship to critically engage with the sociotechnical controversy of AI. By focusing on the case of the AI Commission’s Roadmap for Sweden, this paper explores how emerging futures of AI and education become reasonable. I revisit STS methodologies and inscriptions to trace document practices that emphasise the document rather than the text (Strathern, 2006), showing how temporal politics of foresight technique and affective dimensions in AI chatbots’ vast accomplishments muster fidelity in new allies. The argument posits that the Roadmap’s good practice can be scrutinised through its temporalities (doing it now), effectiveness (doing it well), and uncontroversial assumptions (doing the right thing). This scrutiny demonstrates how specific futures of AI and education take precedence over others as a sociotechnical effect rather than as a powerful technology.