The Shift from AI Models to Agentic Systems - Danila Shtan, Nebius

Watch Danila Shtan, CTO at Nebius, break down the industry's shift from model-centric AI development to agent-based task execution — and what it takes to make agents production-ready. Danila traces how the conversation has moved past model architecture, parameter counts, and knowledge cut-off dates toward what he calls "applied AI": using models to actually get real work done, whether that's handling invoices, managing calendars, or resolving client interactions. He walks through the mechanics of the agentic loop — planning, calling tools, observing results, and retrying when needed — and makes the case that the orchestration "harness" controlling that loop now matters more than raw model intelligence.