As generative artificial intelligence AI (GenAI) tools rapidly enter agricultural extension, policy and innovation narratives increasingly frame conversational agents as scalable substitutes for human advisors. These narratives rest on two flawed assumptions: that digital literacy is an individual skill farmers must acquire, and that AI will reduce the labour intensity of advisory systems. Underpinned by decades of extension research, this session argues instead that advisory work is sustained by collective, relational, and often invisible human labour—and that GenAI risks redistributing and intensifying this labour rather than replacing it.

The session presents findings from action research workshops with extension agents across Malawi, Kenya, Uganda, Nigeria, and Liberia. Through exercises on intent interpretation, problem diagnosis, and decision-making under gendered constraints, we document the forms of cognitive, social, and emotional labour that remain essential even in GenAI-enabled systems. These include sensemaking of incomplete farmer queries, mediation of algorithmic recommendations, trust-building, troubleshooting, and the informal onboarding of marginalised users through peer networks.

Participants will learn how GenAI reshapes advisory labour, where new burdens and responsibility asymmetries emerge, and why these dynamics are often hidden in digitalisation efforts. The session contributes a framework for anticipating persistent human roles in AI-enabled extension and highlights design and governance choices needed to recognise, resource, and protect this labour.

The session aligns with ICT4D themes on responsible AI, digital inclusion, and labour futures, offering a grounded counterpoint to automation-centric narratives and practical implications for designing equitable digital extension systems.