TRIAGE-AI: A clinician-facing implementation-readiness framework for evaluating artificial intelligence in cancer pathways.

Abstract

Artificial intelligence (AI) is increasingly being offered for use in cancer pathways, but clinicians still lack a practical method for judging whether a specific tool is ready for adoption in routine care. Existing standards such as DECIDE-AI, TRIPOD-AI, CONSORT-AI and evidence standards for digital health technologies are essential, but they are primarily directed at developers, trialists and regulators rather than clinical leads making procurement, governance or pathway decisions. This article proposes TRIAGE-AI, a clinician-facing implementation-readiness framework covering six domains: true clinical need, robust medical evidence, inter-site consistency, adoption fit, governance and effective decisions. The framework is intended to complement existing reporting and evaluation standards by translating their outputs into a structured clinical adoption decision. Using urological oncology as a worked example, TRIAGE-AI differentiates applications that are approaching clinical readiness, such as prostate magnetic resonance imaging AI and selected pathology biomarkers, from applications that remain premature or high risk, including renal mass characterisation tools and autonomous large language model decision support. The current evidence supports TRIAGE-AI as a framework developed and applied within urological oncology, with potential for broader application pending further testing. It is offered as a pragmatic method for clinical governance discussions, local validation planning and implementation-readiness assessment.