JMIR Publications

ABSTRACT

Most literature frames artificial intelligence (AI) adoption in healthcare revenue cycle management (RCM) as a binary of automated versus manual work. We propose a staged-autonomy framework that instead models adoption as a progressive transfer of decision authority from humans to AI, developed through a structured narrative synthesis (PubMed, supplemented by policy documents and industry sources, 2022-2026) combined with analysis of publicly documented institutional reforms, and applied to Saudi Arabia's centralized National Platform for Health Information Exchange Services (NPHIES). We define three stages (Augmented, Autonomous-with-Exception, and Full Autonomous RCM), distinguished by the locus of decision authority, with operationalized transition enablers and falsification conditions for four testable propositions. Applied to Saudi Arabia, the framework exposes a governance-data paradox: NPHIES substantially resolves the data-fragmentation barrier that dominates fragmented markets, leaving accountability for autonomous financial-clinical decisions as the binding constraint. The concurrent transition to Australian Refined Diagnosis-Related Group (AR-DRG) casemix payment, licensed government-to-government from Australia, raises rather than lowers the classification and explainability bar, and payer-side analyses link AI-assisted coding to case-mix-intensity inflation, a contested finding whose contestation itself motivates the independent coding-integrity audit we propose. Centralized claims infrastructure is necessary but not sufficient for advanced AI autonomy: regulators should pair explainability and liability standards with a coding-integrity audit independent of the coding AI itself. No jurisdiction yet publishes a zero-human-touch claims indicator, the most useful near-term step for assessing this transition anywhere.