Back to Now Showing

AI Experience

BWW Agent

A beloved chicken wing brand steps in to the world of AI as it navigates its place.

Trailer

The setup

OpenAI had opened agentic ordering for Target and Uber Eats. We set out to put BWW inside ChatGPT so AI native shoppers could build a bag and order through an agent. A concept was first created by my engineering team without any UX support initially. I came in to redesign the proper ordering experience, working entirely inside OpenAI's SDK and agent patterns without using our design system.

Designing in someone else's house

Every interaction had to live inside OpenAI's SDK, so my first job was translating what a BWW order is into GPT native patterns. Rather than force our brand in, I chose the OpenAI patterns that would feel trustworthy to a ChatGPT user.

Cutting checkout

Public distrust of entering card details with an AI, plus OpenAI's guidelines, ruled out in chat checkout. I led a clean handoff instead. The order is built in GPT, then passed to our native site to complete, protecting user trust over owning the last click.

Location before menu

Our menu data is unstructured and only resolves once a location is chosen, so I led with location selection. That guaranteed every price and item shown was accurate to where the customer would order, choosing accuracy over open ended browsing.

Selected scenes

Location first entry
Location first entry
Product Discovery Via Agent
Product Discovery Via Agent
Product Detail Page
Product Detail Page
Bag within agent
Bag within agent

The premiere

Six weeks in, the numbers told a mixed story. Visits and agent engagement were strong, but few orders were completed. The location first gate, the very choice that guaranteed accuracy, drove early abandonment. Timing compounded it with World Cup marketing pulling brand focus and a wave of critical public feedback as other QSRs launched on GPT.

3