The DESIGN MIND is a way to organized specialized, intelligent “Design Agents” into a comprehensive design system able to quickly synthesize complete designs. The current process for large, complex systems such as ships, aircraft and large processing plants is to run a synthesis tool to produce early concept and feasibility designs, then iterate with increasingly detailed subsystem and component designs with weakly-connected specialists in a large design organization. Periodic summary reviews highlight mismatches between the components and subsystems. Eventually, many months or years later, a complete design is produced. The Design Mind approach automates this process to produce complete designs in a matter of hours or days. The key to this process is recognizing a basic pattern between the subsystem and component designers. A “customer” higher in the system hierarchy requests a design for one or more “suppliers” to resolve issues with the customer’s current design instance. The suppliers formulate optimal or near optimal responses to the request and produce a response. Human-in-the-loop suppliers produce good answers to good requests in a matter of days or weeks. The agents in the Design Mind approach cannot produce answers as good as a human, and the agent-based customers cannot produce requests as good as a human design team, but by exchanging request/response pairs in seconds they allow many orders of magnitude more iterations per design instance than human teams. The automated approach is therefore able to explore more of the design space and ultimately come up with better designs. The proposed approach, hierarchical design, is exponentially complex when done with static agents. This presentation explores the artificial intelligence requirements needed to create a dynamic agent based hierarchical design system to avoid the exponential complexity and enable the scale-up required for truly complex systems.