We present Smart Set Assembly (SSA), a multi-agent AI system built to assist set dressing artists in complex 3D scenes. The system was designed around a central question: can AI agents, guided by explicit artistic intent, accelerate set dressing workflows while preserving creative control and production-quality USD scene hierarchy?We built a five-agent orchestration framework, authored over 60 constrained execution functions, and developed a structured prompting methodology that teaches LLM agents to reason about spatial composition rather than simply execute commands. This paper presents our learnings focusing on what worked, what failed, and what surprised us.