This study investigates the integration of artificial intelligence into the Think-Pair-Share (TPS) methodology through a learning analytics lens. Using a mixed-methods quasi-experimental design (N=140), we examined how an AI-enhanced collaborative platform influences creative thinking among computer science undergraduates. The experimental group (n=80) utilized a Google Gemini-powered chatbot for scaffolding, while the control group (n=60) used a standard platform. Through comprehensive learning analytics, we identified optimal AI integration in moderate similarity ranges (0.3-0.7), achieved by 70% of participants. The experimental group demonstrated 30-37% productivity gains and 99% increase in thematic diversity, with moderate lexical standardization effects. Our findings provide empirical evidence for designing educational technologies that balance structured support with creative freedom in AI-enhanced collaborative learning.