The Future of Agricultural Extension and Economics: Integrating Artificial Intelligence, Behavioural Science, Climate Resilience and Inclusive Markets

Abstract

Agricultural extension is being reshaped by four simultaneous pressures: the arrival of low-cost digital and artificial-intelligence-based advisory tools, the reframing of farmer decision-making by behavioural economics, the intensification of climate risk and the restructuring of agri-food markets. These developments are usually reviewed in separate literatures, which obscures how they interact in determining whether advice changes practice and whether practice change raises and protects smallholder incomes. This critical narrative review examines that interaction. Peer-reviewed economics, agricultural, computer-science and climate-services literature published from 2004 to July 2026 was identified through multidisciplinary scholarly indexes, economics repositories, institutional sources and citation tracking, and was appraised with particular attention to identification strategy, external validity, field validation and distributional analysis. The evidence indicates that digitally delivered advice produces modest but cost-effective improvements in adoption, whereas effects on yields, prices and incomes are inconsistent. Image-based diagnostic models achieve high accuracy on curated datasets but lose performance under field conditions, and large language models remain supported mainly by design studies, benchmark tests and surveys of extension staff rather than by evaluations of farm outcomes. Behavioural mechanisms, particularly limited attention, present bias and social learning through well-chosen messengers, are well documented, yet generic nudges and behavioural framings frequently show small or null effects. Climate information services and risk-reducing technologies can crowd in investment, but uptake is strongly conditioned by farmers' resources, gender and trust. Market information alone rarely improves farm-gate prices where intermediaries hold bargaining power, and gains appear larger when market access rewards quality or relaxes liquidity constraints. The review proposes an integrated framework in which advisory technology is necessary but insufficient, and in which risk, liquidity, market structure and social position determine the returns to information. Priorities include field trials of generative advisory systems with safety outcomes, bundled designs that address several binding constraints, evaluation at scale and governance arrangements for agricultural data. The central conclusion is that the future of extension depends less on the sophistication of advisory algorithms than on their integration with institutions that manage risk and distribute market gains.