Between Algorithm and Instinct: A Phenomenological Study of Critical Care Nurses' Decision‐Making in <scp>AI</scp> ‐Supported Care

Abstract

BACKGROUND: Artificial intelligence (AI) is rapidly reshaping critical care through predictive analytics, intelligent monitoring and decision-support tools. While these innovations may enhance early detection and workflow efficiency, they also raise professional questions about transparency, explainability, data bias, accountability and the preservation of compassionate, human-centred care. Critical care nurses, positioned at the bedside where AI outputs are interpreted and enacted, experience these tensions directly, yet their lived experiences remain underexplored. AIM: To explore critical care nurses' lived experiences of clinical judgement in AI-supported care, focusing on how innovation influences professional integrity, ethical accountability and human-centred practice. STUDY DESIGN: A qualitative phenomenological study was conducted at King Faisal University Health Care settings in Saudi Arabia. Semi-structured, in-depth interviews were undertaken with critical care nurses who routinely interacted with AI-supported clinical systems. Data were analysed using reflexive thematic analysis informed by Braun and Clarke's six-phase framework. The study followed the Standards for Reporting Qualitative Research (SRQR). FINDINGS: Sixteen nurses participated. Four interconnected themes were identified: (1) Balancing algorithmic input and professional judgement, where AI was valued as a prompt for vigilance but required contextual interpretation rather than automatic compliance; (2) Instinct informed by experience, describing embodied and situational knowing that nurses perceived as essential when AI outputs did not capture patient complexity; (3) Ethical weight and accountability, reflecting heightened responsibility and concern about scrutiny when following or overriding AI recommendations; and (4) Preserving human-centred care, highlighting deliberate efforts to protect relational nursing roles, patient-family communication and professional identity amid technology-dense workflows. CONCLUSIONS: Nurses experienced AI as transforming the conditions of clinical judgement rather than replacing it. Innovation was welcomed when it supported early recognition and prioritisation, but nurses emphasised that integrity in AI-supported care depends on maintaining professional discretion, ethical accountability and human-centred values. RELEVANCE TO CLINICAL PRACTICE: Implementing AI in critical care should include governance and education that strengthen nurses' critical appraisal of AI outputs, clarify accountability and support transparent, explainable systems. These steps can help ensure AI's impact enhances safety and efficiency without eroding human-centred critical care nursing.