Enterprise AI upgrades usually begin with existing business capabilities, interfaces, and workflows. This talk introduces the ADPS dual-axis design-pattern framework, which organizes patterns by cognitive function and execution topology. Through architecture diagrams and examples—including planning and execution, progress tracking, handoff chains, and approval gates—it shows how to select and combine patterns under real business constraints. A payroll SaaS modernization case and an A*STAR AgentOS demonstration connect these patterns to practical system design and user experience.