Sorting Out Agencies in AI-supported Radiological Work Practices

The article examines how narrow-task artificial intelligence (NT-AI) implementation reconfigures professional practice and agency in radiology departments. Drawing on practice theory and qualitative data from two Danish hospitals, we analyze how radiographers and radiologists interact with NT-AI systems in their daily work. The findings reveal that professional expertise is not simply diluted by AI but transformed through complex negotiation processes involving trust and responsibility. Professionals develop new interstitial expertise by learning when to trust, question, or override AI suggestions. Rather than creating a zero-sum game where AI gains agency at humans’ expense, implementation produces new forms of relational expertise while preserving distinct forms of human and non-human agency. We propose that successful AI integration requires practicecentered design approaches that recognize how agency materializes through relationships between people and technologies, rather than viewing implementation as merely a technical challenge.