Evidence-Gated Autonomy in Healthcare AI Agents: A Systematic Review of Action Scope, Human-Oversight Effectiveness, and Multi-Step Safety Failures

Background: Healthcare AI agents increasingly perform multi-step workflows and tool use, but evidence on their operational autonomy, safety, and human oversight remains limited. This review examines whether greater autonomy is supported by empirical safety and oversight evidence.Methods: PubMed/MEDLINE, Scopus, and Web of Science were searched from January ۲۰۲۲ to July ۲۰۲۶. Of ۴,۰۹۶ records, ۲,۵۵۷ remained after deduplication. Fifty priority reports were selected for full-text review; ۳۴ were assessed and ۲۸ met the inclusion criteria. Data on workflow, architecture, autonomy, safety, oversight, and evaluation were narratively synthesized.Results: Among ۲۸ studies, ۱۸ were published in ۲۰۲۶. Thirteen addressed diagnostic or clinical decision support, nine EHR or administrative workflows, four treatment planning or prescribing, and two patient-facing or cross-domain applications. Autonomy levels were L۰ in five studies, L۱ in ۱۳, L۲ in three, L۳ in one, and L۴ in six; none reached L۵. Most L۴ systems were evaluated only in sandbox or retrospective settings. Only three studies empirically assessed oversight effectiveness.Conclusions: Current evidence supports tool-using and workflow-capable healthcare agents, but not unrestricted clinical autonomy. Greater autonomy should require stronger evidence of safety, failure containment, and effective human oversight.

کلیدواژه ها:

Artificial intelligence ، AI Agents ، Agentic AI ، Healthcare Workflow ، patient safety ، Operational Autonomy ، Action Scope ، Human Oversight ، Oversight Effectiveness ، Multi-Step Failure ، Evidence-Gated Autonomy ، Systematic review