Artificial Intelligence in Bariatric Surgery: A Clinician-Centered Framework for Surgical Readiness, Workflow Integration, and Decision Support

Background Artificial intelligence (AI) shows promise for risk prediction and decision support in bariatric surgery, yet most models are retrospective, internally validated, and insufficiently tested in real-world workflows. Methods This clinician-centered narrative review synthesizes AI applications for perioperative risk prediction, including anastomotic leak, venous thromboembolism, readmission, and long-term weight-loss trajectories. It focuses on the translational requirements for safe surgical decision support, including external validation, explainability, equity assessment, workflow integration, governance, and clinical impact. Results Current AI applications focus on prediction rather than autonomous action. Models exist for complications, readmissions, and weight-loss outcomes, but many are limited by retrospective design, insufficient external or prospective validation, opaque individual risk estimates, sparse subgroup analyses, poor workflow integration, and limited evidence of improved decision-making or outcomes. Discussion Predictive performance alone is insufficient for clinical utility. AI tools must deliver interpretable outputs, perform reliably across patient groups, integrate into bariatric workflows, and trigger actionable care pathways. Without these features, accurate models remain research-stage tools. Conclusion AI is not yet broadly ready for routine bariatric use. Its near-term role is as adjunctive, multidisciplinary decision support within surgeon-led pathways. The proposed readiness framework and checklist may help surgical teams triage AI tools for research-only use, pilot testing, or responsible deployment.