Event-driven Agents: Why your app shouldn't wait for the user to ask

Most agentic apps use the turn-based chat model. But a lot of the real value in agentic apps comes from something happening, not from a user prompt: a flight gets delayed, a new listing hits the market. Forcing that into a request/response loop means the agent can only ever react, and it puts the burden of noticing on the user. This talk argues that chat is a UI, not an architecture. The right pattern underneath is event-driven: an agent that senses a change, decides whether it's worth acting on, and reaches out proactively. We'll walk through real examples, including an airline agent that rebooks a delayed flight before the traveler notices and a real estate agent that surfaces a strong listing midweek instead of waiting for Saturday's open house. You'll walk away with a concrete way to evaluate whether an event was worth surfacing at all, using an LLM-as-judge on the decision itself, not just the message and a map of what to build before you ship an agent that reaches out on its own.

By attending this session, your contact information may be shared with the sponsor for relevant follow up for this event only.