Jun '25 - Mar '26

Shipped Q1 2026

AI Image Generation for
Uber Marketing

AI Image Generation for Uber Marketing

Details

Role

Product Designer via Haus

Focus

Product Design, Strategy, UI, UX, Prototyping

Scope

Owned and designed the end-to-end product experience

A large budget is allocated per
year on external agencies for campaign
images that take weeks to deliver.

A large budget is allocated per year on external agencies for campaign images that take weeks to deliver.

Overview

Marketers kicked off lengthy projects with external agencies for minor changes like swapping a color or localizing an image for a different market. The process involved briefs, kickoff meetings, and multiple rounds of feedback, with turnaround times of 1-2 months. But more often, they settled for existing assets because the timeline couldn't keep up with campaign cycles.

Together with the Uber Marketing team, I helped design and launch 3 features for the 0 - 1 AI image generation tool within the existing asset library to transform how they create brand-compliant campaign visuals.

Marketers kicked off lengthy projects with external agencies for minor changes like swapping a color or localizing an image for a different market. The process involved briefs, kickoff meetings, and multiple rounds of feedback, with turnaround times of 1-2 months. But more often, they settled for existing assets because the timeline couldn't keep up with campaign cycles.

Together with the Uber Marketing team, I helped design and launch 3 features for the 0 - 1 AI image generation tool within the existing asset library to transform how they create brand-compliant campaign visuals.

Some details have been generalized and visuals simplified to protect confidential information.

Outcome

Outcome

Outcome

Reduced image turnaround from weeks to minutes by replacing external agency coordination with self-serve AI generation, enabling marketers to create fresh and tailored visuals for every campaign.

01 Prompt-based generation

Describe your idea with just a few words or upload an existing asset to guide AI and create custom images from scratch.

02 Illustration or Photo Realistic

Choose from brand-compliant styles to transform your visuals and match your campaign convincingly.

03 Image iterations

Refine images through natural conversation to remove a logo, change a color, or resize for a different channel without aligning with an agency.

The style selector

The PM initially proposed multi-select to give marketers more flexibility. I raised a concern that generating multiple styles at once would make outputs harder to evaluate, and engineering confirmed that single-style generation would produce clearer results. I explored three directions and we aligned on single-select with large preview cards to help marketers make confident choices without extra interaction.

Simplify the metadata panel

From five fields to two. I removed system IDs and file format for the system to handle automatically, and relocated inspiration upload into the prompt area.

Designing for brand safety

Brand compliance was a core constraint. Curated style options enforce consistency at the point of creation, and a disclaimer reinforces that all generated images still follow the standard approval process. This gave marketers freedom to experiment without bypassing the trust the brand team needed.

Reflection

The core design problem isn't the input or the output but the gap between them. This project taught me to design the bridge that shapes expectations before generation, builds mental models that hold up when results surprise, and makes the next action obvious when they don't. That's where trust is either built or lost.

More projects

Flex Benefits Card Website, Uber Health

Flex Benefits Card Website, Uber Health

Cachè Color Picker, Hermès

Cachè Color Picker, Hermès

Communication Preferences, Uber For Business, Rides, Eats

Communication Preferences, Uber Rides, Eats

©️ 2026 Claire Wenhsun Li.

©️ 2026 Claire Wenhsun Li.

©️ 2026 Claire Wenhsun Li.