From Brand Voice to Brand Behavior
Session Overview
Every brand has spent the last twenty years perfecting how it sounds: the voice, the tone, the message, the restraint. It was a system you could largely control.
AI changes that. A model does not simply repeat what a brand says about itself. It pulls together signals, decides what matters, resolves contradictions, and forms its own understanding of who you are.
That understanding then shapes how the model represents you to customers. It answers a customer’s question. It decides which three brands get named and which don’t. It will shape how a customer spends their money.
Arpita Kumar has spent six years designing how AI systems behave. She will show what models hold onto, what they flatten, what they lose, and why being visible to AI is not the same as being understood by it.
AI is already representing your brand when you are not in the room. You should know how.
Speaker
Over nearly a decade at Google, including six years focused on model design, her work has spanned Google Assistant, LaMDA, Bard, Gemini, Google DeepMind, and Google Search AI.
Her work focuses on translating nuanced human judgment about how AI systems should respond into frameworks teams can train and test against.
She came to AI through storytelling. She studied film at CalArts and worked in film and television, with work distributed by PBS and writing for Issa Rae.
That background still shapes how she approaches model design: character, language, subtext, and what it takes for a system to speak with a point of view.