The core of your AI product is being designed without you. Engineering and product are shaping model behavior through system prompts, data decisions, and evaluation frameworks. These are design decisions. If designers are not involved, someone else is making them.

In AI products, the experience is no longer a static interface. It is generated in real time by models making decisions on behalf of users. Paz argues that this is the shift that redefines what UX work actually is: from designing interfaces to shaping behavior.

This talk introduces a new UX toolkit built around data, context, and evaluations, the materials that define what models know, how they respond, and what success or failure looks like. Paz shows how designers can move upstream, collaborate more effectively with product and engineering, and bring a user-centered lens to the decisions that shape AI behavior before anyone else has framed them.