Cross-Domain Applications of Artificial Intelligence: A Systematic Review of Models, Data Modalities, and Deployment Readiness

Abstract

Artificial intelligence is applied across a wide and growing range of domain healthcare, agriculture, industry, business analytics, cybersecurity, smart infrastructure, assistive technologies, education, and sustainability, yet cross-domain synthesis of AI evidence has lagged behind domain-specific development. A systematic cross-domain review must classify studies not only by application area, but also by data modality, model architecture, decision-support function, deployment pathway, and deployment-readiness concern, to reveal the structural patterns and gaps that single-domain reviews cannot expose. This review presents systematic-style evidence mapping analysis of corpus using a seven-axis taxonomy: application domain, data modality, model family, decision-support function, deployment pathway, deployment-readiness concern, and evidence role. Seven application domains are synthesized, healthcare and biomedical AI, human-centered and assistive AI, agriculture and sustainability, industrial monitoring, smart infrastructure and IoT, business and enterprise analytics, and cybersecurity and distributed intelligence, across ten model families from conventional ML through vision transformers, graph neural networks, Bayesian physics-guided models, generative AI, and federated learning systems. Synthesis reveals that model architecture diversity has advanced substantially, while deployment-critical properties, validated explainability, privacy-preserving inference, robustness, human oversight, and governance-aligned reporting, remain inconsistently addressed. A ten-direction research agenda identifies the most consequential future priorities for cross-domain deployable AI. Note that full PRISMA reporting would require explicit database-search records, screening procedures, and eligibility criteria beyond what this curated corpus provides.