Athleta | AI Assistant
My Role
UX/UI Direction LeadProduct Lead
Product Strategy
Type of Work
UX/UI Design
Product Strategy
AI Strategy
Prototyping
Branding
Research
Background
​
Many apparel brands have tried Stylist "Chatbots" over the years, starting most notably with Levi’s Ask Indigo and Tommy Hilfiger. These early chatbots would give rather generic recommendations and were mainly geared towards helping customers sort/filter through large inventories. For example, one of the main demonstrations of Ask Indigo focused on helping customers find “light-blue jeans under $50” (a combination of price and color filtering). This brought back a subset of their catalog filtered by these two criteria.
​
More recent chatbots allow unscripted, more conversational interactions made possible through advances in Natural Language Processing (NLP, which allows customers to interact using a more conversational chat), and AI that can leverage, for example, (1) customer purchases or preferences, (2) keywords from customer reviews (I.e. “wedding” or “black-tie”), (3) and even non-Gap data sources.
​
The goal of this project, the first in our AI roadmap, was to launch an AI-driven Virtual Store Associate with Athleta.
Discovery
Our team spent a few days each in Athleta stores across the country observing Athleta sales associates and trainers, interviewing customers, and in some cases doing some serious shopping.
​​
We also reviewed sales training material and interviewed corporate SMEs, all to undersand the primary interactions we needed to support, and to ensure that that our virtual associate could talk and act like an real Athleta associate.
While we worked on the overall UI for the AI Stylist, the lions share of the design phase was spent on the voice / tone, and our recommendation engine. ​
​​
Voice / Tone: Generative AI typically comes off as being smarmy ("That's a fantastic choice"), while the Gap brands are generally more straightforward.
​
​
Design
Recommendations: One of our product tenets was "Don't look stupid". We really sweated the logic and cleaned up the product data to ensure if you asked for leggings you got leggings (not pants). Don't put the campfire in the tent.
After three months we launched the Atheta AI Assistant with a series of internal-only dogfooding trials. Dozens, then hundreds of Gap employees searched for marathon training gear, gifts for their family, or asked us funny questions to make sure we weren't answering math equations or riddles. We also conducted user studies to gather feedback on the service, and got ready for launch.
Deliver
Key Learings from dogfood + user research
The full vision for a Gap-wide assistant was to offer suggestions throughout each brands' customer journey (shown here customized for BR). We developed a full roadmap to offer sizing suggestions, style advice, answer customer support questions... Whatever the customer needed, wherever they needed it.
​
In addition, the content proved extremely useful to store associates that we kicked off another project to support associates through their in-store mobile app.
Future vision
