AI-First Creative Agentic Workflows
As Uber’s organizational brand needs scaled, it became clear that traditional, creative-only workflows and siloed sources of truth were creating bottlenecks.
Our Brand Experience team faced a critical challenge: expanding beyond foundational brand guidelines to establish a proactive, AI-first point of view that empowered teams across the company. To solve this, I led the development of a suite of brand-led creative agents—utilizing custom GPTs and Google Notebook—to transition our operations from manual, on-demand consultation to scaled, self-service AI assistance.
Empowering the Enterprise
We developed an expanding set of GPT and LM reference agents designed to execute ad-hoc, on-brand solutions across multiple disciplines:
Always-On Brand Intelligence (Google Notebook): We launched live, AI-first knowledge bases for Uber’s Brand Architecture and Editorial Style Guide. This empowered non-creative teams with immediate, chat-based resolution for brand knowledge and copy standards.
Visual Generation & Prompting (Custom GPTs): To ensure visual consistency, we built a Global Photography Prompter that generates highly detailed, on-brand photography and illustration prompts for generative AI tools. Additionally, we developed an experimental Brand Layout GPT, leveraging ChatGPT’s Image Gen 2.0 to explore variable, robust on-brand concept layouts.
Copy Assistant (Custom GPT): To further democratize creation, we initiated the development of a Copy Assistant to provide immediate, automated guidance for on-brand voice authoring.
Business Impact & Operational Efficiency
By extending the consultation services of the central Brand team directly into self-service AI tools, we established an ecosystem that transforms fragmented knowledge into an automated engine for consistent, scaled content creation. This AI-first infrastructure was designed to fundamentally shift how work gets done, with projected impacts estimating a 30% decrease in ad-hoc brand consultation requests, a 75% reduction in brand inaccuracies, and a 30% increase in non-Creative Team assisted output.
