Traditional ease of use evaluations are too slow for agile B2B product development. We built and tested "AI UXpert," a generative AI agent, against 20 human experts. Our study revealed AI UXpert is faster and stricter but struggles with contextual issues. A crucial lesson is AI UXpert’s unique value as a "cognitive outsider": it identified a systemic risk all 20 experts missed. Practitioners value AI UXpert’s speed and unique insights but remain cautious about its credibility, pointing to a trust gap. Our findings advocate for a human-AI partnership where AI handles rapid screening, freeing experts for deep, strategic analysis. This collaborative vision provides a roadmap for effective AI integration in user experience design.