Do you know how much your AI system spends when someone just says “Hello”?

Many AI projects start as a simple “ask me anything” bot: it greets, responds, and every interaction feels magical — until the bill arrives. This session will make the hidden cost of “Hello!” visible and show how to avoid paying the “big model price” for every small request.

We will start with a simple CLI chatbot and explore how smarter routing can decide whether a message needs a strong model, a smaller model, a predefined response, or no model at all. Along the way, we will look at guardrails, fallbacks, cost tracking, latency, and practical ways to make AI systems safer and more efficient.

Attendees will leave with a practical way to think about building AI systems that are not just smart, but also cost-aware, safer, and easier to operate.