Search is undergoing its biggest architectural shift since the rise of neural retrieval. This talk traces the evolution from classical IR pipelines to next-generation search systems powered by LLMs and agentic AI. We’ll walk through the modern industrial search stack from query understanding and retrieval to ranking, reranking, and answer generation and show how each stage is being reshaped by foundation models. You’ll see how LLMs reduce brittle rules in query understanding, expand long-tail recall through semantic reformulation, strengthen coverage via hybrid retrieval, and introduce real reasoning signals into ranking. We then explore the emerging agentic layer: systems that plan, call tools, refine queries, and self-evaluate to deliver more adaptive and user-aligned results. Alongside the promise, we’ll examine the engineering realities: latency, determinism, orchestration overhead, and quality measurement. Attendees will walk away with a clear end-to-end mental model of what “LLM-native search” looks like today, what agentic search could become, and how to pragmatically bridge the gap from ranking-based systems to reasoning-driven ones.