Operational and Financial Evidence for Generative AI in United States Healthcare Administrative Workflows

Abstract

Generative artificial intelligence (GenAI) is increasingly used for clinical documentation and administrative work, but reported benefits often mix time savings, capacity, billing proxies, and realized financial value. This structured narrative review synthesized peer-reviewed evidence on operational and financial outcomes in United States healthcare administrative and hybrid workflows and developed a governance-integrated evaluation framework. PubMed, PubMed Central, publisher records, backward citation searching, and DOI or PMID verification were used to identify empirical studies and relevant syntheses. Fifteen primary studies and one systematic review met the use-case criteria; ten primary studies evaluated ambient documentation, including two randomized trials. Ambient tools generally produced modest reductions in documentation time or work burden, whereas visit volume and productivity effects were inconsistent. One controlled study estimated 1.81 additional relative value units and 0.80 additional encounters per physician-week but did not measure complete implementation costs or net collected cash. Patient-message tools showed utilization near 20% and no consistent time savings. Summarization and coding studies identified clinically important omissions, inaccuracies, and poor agreement with human coding. Evidence of net return on investment remained sparse. Adoption decisions should pass sequential gates for technical validity, workflow fit, operational effectiveness, financial conversion, and equity and sustainability, with total cost and human review measured before return on investment is claimed.