Federated Learning (FL) has become a vital method in machine learning, providing strong solutions to HR’s data security challenges. Health-related employment data becomes more sensitive when it comes to health services 5.0; it is necessary to choose the right FL architecture. As centralised and decentralised FL models, they belong to the HR system; this chapter is completely associated, which emphasises privacy, operational efficiency and their effect on regulatory compliance. In centralised designs, model updates from scattered customers are gathered by a coordination server, providing easy control and simple model convergence. However, this strategy can increase the risk of single points and risk of data leakage, such as weaknesses. On the other hand, decentralised FL design, spreading training between colleague nodes, improves fault tolerance and maintains the data area. However, this paradigm can withstand low scalability problems, inconsistent model updates and communication costs. The chapter considers the effect of these architectural paradigms on the workforce, result management and talent collection among other essential HR functions. The legal and moral implications of FL distribution in global companies interact with various data protection frameworks, especially when assessed. The study also looks at the hybrid model, which aims to combine the best features of both architectures to provide a good answer to the complex HR environment. This chapter provides a forward-looking pattern for intelligent, privacy-conscious and adaptable HR systems by integrating FL-techniques with the principles of health services 5.0: Adaptation, moral rule and flexibility. This ends with suggestions for scholars and doctors to use FL to promote human-focused innovation in the workforce.