While working on a vibe-coding project, I found my discovery interviews getting stuck surprisingly fast, really after just a few conversations, I wasn’t really learning anything new. The issue wasn’t volume, but my ability to process what I’d already heard, identify what was still unclear, and adapt in real time. So, naturally, I did what everyone on LinkedIn says and built an AI agent. In this talk, I’ll show how it helps me prepare for interviews, debrief patterns across conversations, and even get live guidance mid-interview and why the agentic layer matters: not just summarizing, but actively identifying gaps, suggesting next questions, and turning interviews into a continuous learning system. You’ll leave with a practical approach to building your own AI-powered discovery workflow.