Real-World Challenges of Intelligent Clinical Workflows

The rise of Agentic AI and LLMs has led to new possibilities for intelligent, adaptive and context-aware healthcare systems. The framework  has been proposed, which involves integration of agentic reasoning, inference enhanced by the use of retrievals, interoperable FHIR representation, and clinician guided planning for the purpose of supporting dynamic, patient-centered workflows. Despite obvious advantages in terms of automated reasoning, patient monitoring and personalized decision-making, implementation of the system in the clinical environment is not free from specific challenges. This paper provides critical discussion of limitations and barriers involved in deploying the framework in clinical practice. In addition, issues regarding regulatory compliance, ethical responsibility, clinicians' trust, explainability and automation bias will be discussed. The main conclusion of this study is that the development of next generation intelligent healthcare systems represented by an intelligent framework has a good potential but still needs resolving deployment-related challenges in regard to safety, interoperability, accountability and scalability to become truly clinically usable.