AI-Enabled Resilient Systems: Healthcare, Sustainability, Cybersecurity, Energy, and Organizational Decision-Making

Abstract

Resilience, the ability of a system to support continuity, adaptation, risk awareness, and recovery under uncertainty or disruption, has emerged as a central requirement for AI-enabled decision systems across critical domains. Healthcare systems must sustain diagnostic accuracy and patient privacy under data constraints and clinical heterogeneity. Energy and infrastructure systems must maintain monitoring reliability under sensor drift and communication instability. Cybersecurity systems must protect data and detect threats in adversarial environments. Agricultural and sustainability systems must function in variable field conditions with constrained hardware. Business and organizational systems must remain agile under economic volatility and pressure of governance. Human-centered AI systems must preserve accessibility, personalization, and ethical oversight for vulnerable users. This structured critical review synthesizes a curated corpus using a seven-axis resilience taxonomy encompassing resilience domain, resilience function, data modality, architecture family, system layer, deployment concern, and decision-support level. Seven resilience domains are examined and mapped across architecture families from conventional ML and CNNs through vision transformers, graph neural networks, physics-guided Bayesian models, generative AI, and federated privacy-preserving systems. Synthesis reveals that while AI capability has advanced substantially, resilience-specific properties robustness, uncertainty quantification, privacy-preserving collaboration, post-deployment monitoring, and governance accountability—remain inconsistently addressed. A twelve-direction research agenda addresses these gaps with actionable future directions and evaluation requirements for AI-enabled resilient systems.