From prediction to practice: artificial intelligence as an enabling layer in liver transplant care - reflections on...

Liver transplantation has become increasingly complex and data-intensive, demanding rapid, high-stakes clinical decisions as patient and organ conditions evolve. The surge of multidimensional clinical datasets has sparked strong interest in artificial intelligence (AI) and machine-learning (ML) as decision-support tools. The 2025 International Liver Transplantation Society (ILTS) annual meeting in Singapore highlighted that AI is now permeating every facet of liver transplantation, from donor-recipient matching and intra-operative navigation to post-transplant pharmacologic management. Early-stage models have shown impressive predictive performance in risk stratification for rejection, infection, and postoperative complications, as well as in individualized immunosuppression dosing and AI-assisted imaging/histopathology. However, real-world implementation has exposed persistent obstacles: data silos, limited generalizability, suboptimal explainability, and workflow friction that can erode clinician trust. Poor integration into existing clinical workflows may also increase cognitive burden for clinicians and reduce adoption of otherwise promising tools. This perspective synthesizes the ILTS 2025 messages and recent transplant literature, proposing a pragmatic three-stage roadmap for AI adoption over the next five years. The roadmap moves from near-term, well-validated decision-support alerts integrated into electronic health records, through mid-term hybrid human-AI teams that embed explainable AI (XAI) into clinical workflows, to long-term global AI consortia that foster federated learning and equitable model deployment. Central to this vision is a shift from "AI as an autonomous decision-maker" to "augmented intelligence that amplifies human expertise while remaining transparent and equitable," ensuring that AI serves as an enabling layer that augments, rather than replaces, clinical judgment.