The shopping agent was built to provide reliable product recommendations by ensuring every suggestion meets the user's requirements, including price, category, and availability. Through competitive evaluation against standardized shopping tasks, the team discovered that a structured pipeline—understand, search, filter, verify, compare, and recommend—outperformed relying solely on a powerful language model. The model was reserved for tasks requiring judgment, while predictable operations were handled through explicit processes. This approach achieved first place on the public leaderboard with a 92.74% success rate and reportedly generated approximately $200,000 in competitive performance returns. The key lesson is that effective workflow design and rigorous verification are more important than model capability alone when building reliable AI agents.