Intelligent Agriculture for Plant Disease Management: A Critical Review of Operational Value, Field Reliability, Adoption Barriers, and Research Priorities

Plant disease management increasingly integrates artificial intelligence (AI), Internet of Things (IoT) sensing, remote and proximal sensing, disease forecasting, robotics, cloud–edge computing, and decision-support systems (DSS). This critical narrative review evaluated their operational value, reliability under production conditions, barriers to adoption, and priorities for improving field performance. An initial 99-record bibliography was screened, yielding 45 eligible sources; targeted searches of DOI/indexing services and publisher databases from 2010 to July 2026, while retaining earlier foundational studies, identified 35 additional DOI-verified sources, producing an 80-reference corpus. Evidence was appraised across five deployment-oriented domains: field representativeness, diagnostic ground truth, comparator quality, uncertainty and failure reporting, and operational outcomes. Intelligent tools provide three main benefits: earlier or more scalable detection, continuous and spatially explicit risk assessment, and targeted intervention. Field evidence is strongest for established disease-warning systems and selected precision-spraying applications, whereas image classifiers, foundation models, and autonomous robots remain more commonly validated in curated or crop-specific settings. Distribution shift is a major reliability concern, with commercial-field studies showing substantial performance losses under open-set conditions. Wider adoption is constrained by interoperability, cost, maintenance, connectivity, data governance, trust, liability, and unequal access. The most credible near-term pathway is supervised, context-sensitive autonomy. Future progress requires multi-farm and multi-season validation, explicit uncertainty reporting, FAIR (Findable, Accessible, Interoperable, and Reusable) data architectures, economic evaluation, climate-adaptive hybrid models, and end-to-end comparisons with best conventional practice. Intelligent agriculture should therefore be developed as accountable plant-health infrastructure that keeps plant pathologists, advisers, and farmers in the decision loop.