Disrupting the shopping experience
How Artificial Intelligence can launch eBay into the future of eCommerce.
Client
eBay
Services
iOS app design, product design, workshopping, UX & wireframing, prototyping.
Project Overview
Develop an ambitious vision for the future of AI within eBay and kickstart the pipeline of valuable ideas that can launch immediately. Our aim was to identify and develop ideas that establish eBay as a disruptor in this space, create strategic value, and design an iterative feedback loop.
Roles & Responsibilities
Lead a team of 18 people in a hackathon-type sprint, taking them from conception to deployment. Every day myself and a strategist generated new workshops to help our team understand the current task, surface wants and needs from the customer in relation to our section of the experience, generate a variety of creative solutions, and group those solutions to form concepts. From there I managed the design and development teams as we worked simultaneously to push those concepts and build out a high-fidelity prototype and a fully functional demo.
Problem Statement
Exploration and evaluation can be overwhelming on eBay, even for Enthusiasts. There’s so much to sift through and no easy way to quickly assess and compare items based on unique preferences or real-life scenarios. It’s hard to tell what’s ‘right for you’ when the current solution is built for everyone.
Opportunity
Our vision is to leverage generative AI to super charge search and make the browse experience more conversational, contextual, and fun.
With eBay.AI, buyers have an always-available, smart shopping Agent that meets them where they are, responding in real-time with the best items, easy-to-scan info, contextual prompts, and more.
Scope of work
We focused on the Agent’s foundational anatomy and core capabilities, to bring it to life at speed. This includes who the agent is, where it operates, and how it communicates.
Who is the agent?
We intentionally chose a middle-ground on the spectrum of Agent personification, mirroring eBay voice and tone. With testing, the Agent’s identity could scale to be more personalized.
Where does the Agent operate?
Today, the Agent is on-platform though we have off-platform aspirations. We explored a range of dedicated and contextual entry points to optimize Agent engagement, based on page-specific user needs and jobs to be done.
When does the agent communicate?
Likewise, we explored if and how buyers could “summon” the Agent, and defined when the Agent might proactively engage buyers, based on need.
Anatomy overview
The eBay.AI Agent is designed to be easily activated when you want it, and surfaced when you need it. Having a variety of entry points and activation thresholds allows the Agent to thrive throughout a customer’s journey with personalized and contextual prompts.
Project timeline
CX Immersion
Research the current customer experience, pulling from existing studies as well as evaluate the current competitive ecosystem.
Surface a variety of wants and needs in relation to our specific customer experience.
Workshop and generate creative solutions on how to solve these wants and needs, ultimately forming concepts.
AI Immersion
Onboard the team onto the powers of AI and it’s vast capabilities, agnostic of an e-commerce experience.
Workshop a variety of ways in which AI can benefit the current customer’s shopping experience based on the wants and needs we addressed in the previous week.
Begin to map out a new CX with the addition of our AI solutions.
Design & Delivery
Deliver refined design concepts, UX flows, and fully animated prototype.
Deliver a working demo for how our AI concept can be implemented into the current eBay ecosystem.
Personification & presence
We reviewed the smart assistant landscape and mapped a spectrum of potential agent identities, considering how and where the Agent shows up for users. We intentionally chose a middle-ground on the spectrum of personification, mirroring eBay voice and tone.
We also explored a range of dedicated and contextual entry points to optimize Agent engagement. Today, the Agent is on-platform though we have off-platform aspirations.
Entry points
Users can access the agent from practically anywhere within the app, making it intuitive and easy to get the assistance they need where and when they need it.
Activation threshold
The Agent is able to proactively engage users based on their needs and certain triggers. This ensures users don’t get stuck and can get the relevant help and pivots needed to accomplish their goals.
Conversational frameworks
Our frameworks are based on four key “search intents” that the Agent’s AI has been trained to recognize and respond to.
Natural Search
If the user intent is recognized as ‘Search,’ the Agent will optimize the user's prompt into search queries and display search results.
General Knowledge
If the intent is ‘General Knowledge,’ the Agent will provide interactive choices with categories for the user to drill down. The Agent will then optimize the user selected category to search queries to display product results.
Expert
In the “Expert” intent, the user is looking for instructions on fulfilling specific tasks. The Agent provides relevant instructions along with interactive choices for the user to drill down. The user can even dive further in with additional questions to see relevant product results on the same topic.
Discover
If the intent is ‘Discovery,’ the Agent responds by displaying relevant product categories (as images and concepts) based on the users prompt (e.g., dresses, bags). The user can select the styles they want to further explore. If the category is not supported, a prompt will be treated as a ‘Natural Language’ intent.
Concept definition
Design, Product, and Engineering collaborated to develop a fully functional coded demo serving as a proof of concept. This experience showcased the best of the Agent’s capabilities focusing on four conversational scenarios (Search, General Knowledge, expert & Discovery). For each interaction eBay.AI identifies the intent, queries, concepts and categories and provides a conversational, interactive response that users can further explore or refine product results.
Concept 1
Personal shopper
Now, you can browse, bid, and buy with the ease of chatting. The eBay Agent understands conversational search queries to discover the thing you need next; whether it's a gift for your 5-year-old niece, the best poles for your new skis, or parts for your restoration project. Our Agent has your back before you buy with alerts about price drops, handles bidding and purchase, and will support you post-purchase too, with real-time tracking and updates.
Concept 2
Smart researcher
The eBay Agent helps you compare items with easy-to-scan summaries powered by what’s important to you. From resale value and condition, to shipping speed and carbon footprint, you’ll get results in an instant, so you can find your new fave thing. And at every visit, we’ll have a wrap-up of things you may have missed to pick up where you left off.
Concept 3
Scan to shop
With the eBay Agent, what you see can be yours in an instant. Live-scan something IRL, snap a photo with the app, share a screenshot from your social feed, or text eBay directly to find what you’re looking for. Buy it now, or save it for later in your virtual closet or living room. The agent will even help you flip it when you’re ready to part ways.
Research approach
We ran an unmoderated concept test with 15 buyers to understand their needs and pain points surrounding the current in-app experience at key touch points. This included capturing buyer considerations and intents surrounding the eBay.AI Agent at various touch points as well as their current relationships to and experience with AI assistants.
Qualitative Summary
Prompt
”What do you think of the eBay agent proactively recommending items and engaging you in your search?”
Prompt
”What do you think about the physical placement of the agent on each of these pages?”
Research Findings
Research validated that the eBay.AI agent can enhance the search and browsing experience of buyers in three key areas.
Clear value proposition of eBay.AI
Value prop is clear in using AI to surface trends, identify trustworthy reviews, improve delivery transparency, confirm item condition, and improve listing photos.
Relevant placement of eBay.AI entry points
The placement of the eBay agent at different touch points landed well with participants. Several suggested improving discoverability of the agent on the homepage with improved contrast compared to nearby filter buttons.
Willingness to try eBay.AI within Search
Participants expressed openness to using AI in initial / broad searches, but need more convincing of the agent’s competence and accuracy before using the agent to dive into item specifics.
Research Opportunities
Insights grounded from buyer feedback paved directions to build a more magical shopping experience through eBay.AI.
Humanize, not sterilize
Injecting personality in eBay’s conversational prompts may prevent eBay from coming off as too plain or robotic.
Differentiate AI experiences from ads
Buyers want AI to look out for what’s best for them and not get pressured into buyer’s remorse.
Build for a need for speed
Buyers lament at how slow and inaccurate Siri and Alexa are at sharing results and would use eBay’s AI if it sped up search.
These AI experiences had to be strategically impactful — result in large scale, meaningful consumer behavior change (eg 10x shopper experience, take back commerce search), novel & magical — “only eBay can,” and near term — can launch in the next quarter and announced in upcoming earnings.
With so many modern apps on the market that tout how seamlessly you can collaborate with coworkers, it was refreshing to get back to the basics of whiteboarding. After the first week of concepting we dove right into storyboarding and sketching, pushing the boundaries of what this experience could be.
