Entrepreneurship is a critical driver of innovation, yet women remain significantly underrepresented in the field due to persistent gender biases that restrict access to resources, funding, and networks. This paper presents insights from developing a serious game that combines decision-tree narratives with Large Language Models (LLMs) to simulate entrepreneurial scenarios and provide real-time bias detection and feedback. Players navigate interactive dialogues that reflect common gender biases in entrepreneurship, fostering awareness and equipping them with strategies to handle discriminatory situations. The game integrates a domain-specific knowledge base of over 20 biases with structured LLM analysis to ensure accurate detection and tailored feedback. Developed through an iterative co-design process involving women entrepreneurs, consultants, and gender experts, the game emphasizes emotional safety and contextual relevance. Our findings from user testing and expert evaluations demonstrate how this approach fosters resilience and preparedness. This research contributes to gender equity in entrepreneurship and AI-assisted educational tools, advancing inclusive and bias-aware innovation ecosystems.