Responsible Artificial Intelligence Off-Boarding in Radiology: Staff Perspectives on Decommissioning and a Proposed...

Area of Science:

• Artificial Intelligence in Healthcare
• Clinical Workflow Optimization
• Health Informatics

Background:

• Current AI governance frameworks prioritize implementation over withdrawal.
• There is a lack of guidance on decommissioning AI tools due to abrupt changes in funding, contracts, or strategy.
• Understanding staff experiences during AI tool withdrawal is crucial for responsible AI lifecycle management.

Purpose of the Study:

• To examine staff experiences following the withdrawal of a chest radiograph AI triage tool.
• To develop a practical framework for responsible AI off-boarding.
• To identify the impact of AI tool decommissioning on clinical workflow and staff burden.

Main Methods:

• Anonymous staff survey conducted two months post-decommissioning of an AI chest radiograph triage tool.
• Survey included questions on workflow, patient benefit, emotional burden, and future AI engagement, comparing with earlier implementation surveys.
• Quantitative data summarized descriptively; free-text responses analyzed deductively using the Job Demands-Resources (JD-R) model.

Main Results:

• Perceived patient benefit from AI remained stable post-decommissioning (70%).
• Logistical burden post-decommissioning was higher than late implementation (35%) but lower than early implementation.
• Staff reported loss of clinical value/efficiency, operational relief, and emotional labor; reporting staff had low response rates.

Conclusions:

• AI tool decommissioning results in both operational relief and perceived clinical loss, with role-dependent effects.
• A three-phase AI off-boarding protocol (pre-withdrawal assessment, graduated transition, post-withdrawal support) is proposed.
• Responsible AI governance must encompass the entire AI lifecycle, including decommissioning.