Industrial systems are undergoing accelerated transformation due to the convergence of artificial intelligence, cyber-physical systems, and modular software design, particularly within the vision of Industry 5.0. This shift increases uncertainty, making rigid solutions unsustainable, especially for small and medium-sized enterprises, requiring faster, more flexible design and integration. Simultaneously, the maturity of multi-agent systems and large language model-based agents is redefining possibilities for intelligent, AI-augmented design processes and context-aware software architectures. As such, this paper explores collaborative architecture design between humans and AI agents as a foundation for evolutive manufacturing concepts, where modular components interact in a logic of system of systems through autonomous, agent-based coordination. The authors emphasize architecture description methods and artifacts, human-AI collaboration, agentic traceability, and the potential of industrial data spaces. The objective is to present a brief survey on these topics, supporting scalable and resilient industrial systems that harness the complementary strengths of human and artificial agents.