AI often feels simple: you write a prompt, you get an answer. But behind that interaction, multiple systems are running, consuming resources, and generating real cost.

In this session, we go under the hood of AI agents and Copilot experiences to understand what actually happens when a prompt is executed. We break down how tokens, queries, pipelines, and Microsoft Fabric capacity contribute to the final cost, and how these costs accumulate at scale.

Through practical scenarios, we compare different approaches, from simple prompts to more complex data-driven questions, showing how design and architecture decisions directly impact cost and performance. We will also discuss common misconceptions and where teams tend to underestimate expenses.

By the end of the session, you will have a clear mental model of how AI costs are generated, how to translate usage into real money, and how to design more efficient and predictable AI solutions in production.