Responsible Artificial Intelligence Off-Boarding in Radiology: Staff Perspectives on Decommissioning and a Proposed Withdrawal Framework
Journal of the American College of Radiology, 2026
Most AI governance in radiology focuses on getting tools in the door, validation, procurement, deployment, monitoring. What happens when a tool is taken away has barely been studied. This paper looks at exactly that: a chest X-ray AI triage tool at an NHS trust was decommissioned in 2025 when pilot funding ended, with no staged transition or replacement. Two months later, staff who'd already been surveyed three times during the tool's rollout were surveyed again.
The results were mixed rather than simply negative. Support for AI-assisted triage stayed high, and most staff still felt it had improved patient care. But logistical burden actually rose again after withdrawal, and comfort with AI reading one's own chest X-ray dropped to its lowest point across all four survey waves. Reporting staff described missing a "safety net"; other radiographers described relief from constant interruptions and difficult patient conversations. The authors use the Job Demands-Resources model to explain this: the same tool had functioned as both a burden and a support, and removing it did neither uniformly.
Based on these findings, the authors propose a three-phase AI Off-Boarding Protocol: pre-withdrawal assessment, graduated transition, and post-withdrawal support, arguing that decommissioning deserves the same structured planning as launch.