Balancing automation and augmentation in AI-enabled decision-making: A business model perspective from healthcare sector

Purpose The increasing integration of artificial intelligence (AI) into organizational decision-making processes raises critical questions about whether decisions should be automated or augmented by human expertise. While prior research presents automation and augmentation as points on a continuum, less is known about their interplay. This study examines how organizations can balance decision-making automation and augmentation in data-intensive, high-stakes environments. Design/methodology/approach Drawing on a theoretical framework integrating paradox theory and business model thinking, we conduct a single-case study of a Finnish intensive care unit (ICU) using AI to support decisions related to acute respiratory distress syndrome (ARDS). We analyze how AI reshapes decision-making through changes in task content, sequencing, timing, and governance. Findings The automation–augmentation balance is shaped not only by task complexity but also by the sequence of tasks, the governance level involved and timing of tasks. We also identify three paradoxes in balancing task automation and task augmentation across different levels of healthcare. Research implications This study shows how AI-enabled decision-making can be understood as a configurational challenge involving temporal and structural dimensions, extending paradox theory and advancing research on AI-driven business models. Practical implications The study offers a diagnostic approach for managers to assess when and how to deploy AI. By considering decision timing, task interdependencies, and governance structures, organizations can optimize resource use, mitigate risk, and enhance resilience in AI-intensive contexts. Originality/value This study bridges the automation–augmentation debate and business model literature. It offers a dynamic view of AI's organizational impact and reveals how human–machine interactions can be structured to support digital value logic.