01 · Lyzr AI · 2026

Zenni

E-commerce experience with an AI-powered conversational shopping agent

Conversational AI
E-commerce
Human–AI interaction
Product UX

Overview

I designed Zenni, an AI-powered conversational shopping agent integrated into the Zenni storefront. Zenni helps users discover products, make product decisions, configure their purchase, complete checkout, and track their order through one connected experience.

The problem

Shopping for eyewear can involve multiple decisions around style, fit, prescription, lens options and price. Traditional e-commerce makes users navigate catalogs, filters and separate product flows to complete these decisions.

My role

Designed the conversational shopping experience end-to-end, including the AI interaction model, conversation patterns, product recommendations, decision-making flows, shopping actions and the transition between conversation and the existing Zenni experience.

Context

Zenni was designed as an AI layer within the existing Zenni e-commerce experience rather than as a separate shopping destination. The goal was to make complex shopping decisions easier while keeping the familiar website experience available.

Challenge

Conversation vs. shopping

Chat is useful for understanding intent, but shopping still requires visual comparison and structured decisions.

Trust in recommendations

AI recommendations need to feel relevant and understandable rather than arbitrary.

Complex decisions

Eyewear involves multiple product and lens decisions that should not feel overwhelming.

User control

The agent should guide users while keeping important decisions and actions in the user's control.

Approach

01
Mapped the existing shopping journey and identified where conversational guidance could reduce friction.
02
Designed Zenni as a shopping interface, combining conversation with product cards, recommendations, quick actions and structured UI.
03
Used contextual questions and guided choices to help users move through complex decisions.
04
Extended the agent beyond discovery into configuration, checkout and post-purchase tracking.

Key UX decisions

Problem: Text-only replies cannot show product comparison, which is essential for choosing eyewear.
Decision: Rendered product recommendations as visual cards inside the conversation.
Why it mattered: This maintains conversational speed while restoring visual comparison shoppers need.

Problem: Recommendations without explanation feel arbitrary and reduce trust in the agent.
Decision: Paired each recommendation with a plain-language rationale tied to what the user said.
Why it mattered: Users can immediately judge if the agent understood them and correct it if needed.

Problem: Pure conversation cannot handle complex configuration tasks like prescription and lens selection.
Decision: Integrated structured UI components for configuration decisions within the conversational flow.
Why it mattered: Important purchasing decisions require clarity that conversation alone cannot provide.

Problem: Context gets lost as users move between conversation and shopping workflows.
Decision: Maintained user context across the entire journey from discovery through post-purchase tracking.
Why it mattered: Continuity keeps users engaged and reduces friction in complex shopping tasks.

Problem: The agent will sometimes misunderstand or make incorrect recommendations.
Decision: Designed explicit repair patterns as visible controls so users can refine, narrow, or restart.
Why it mattered: One-tap recovery keeps failure from ending the session and maintains trust.

User flow

01
User intent
Described in natural language
02
AI understands request
Clarifies one variable at a time
03
Product discovery
Catalog results rendered in-thread
04
Recommendation
Shortlist with plain-language reasoning
05
Product selection
Choose and configure the product
06
Configuration
Lens options and customization
07
Prescription
Prescription entry if needed
08
Checkout
Complete the purchase
09
Payment
Process payment
10
Order confirmation
Confirm the order
11
Order tracking
Track the shipment

Zenni guides users through the entire shopping journey from discovery to post-purchase support.

Design system

Reusable conversational patterns form the foundation of Zenni's interactions.

Agent and user messages: Clear distinction between agent and user communication.
Clarifying questions: Contextual questions to guide decision-making.
Product cards: Visual product representations within conversation.
Comparison patterns: Side-by-side product comparison in-thread.
Quick actions: Contextual shopping actions available in messages.
Configuration components: Tools for customizing purchase options.
Error and recovery states: Clear guidance when understanding fails.
Shopping actions: Add to cart, purchase, and tracking actions.

Final experience

Conversational entry inside the storefront

The launcher sits within the existing site rather than over it, so the AI reads as an assistant to shopping — not a separate destination.

Intent to shortlist
A vague request becomes a small, reasoned shortlist instead of a filtered result page with hundreds of items.

Comparison inside the conversation
Frames are shown side by side in-thread so the decision does not require leaving the conversation.

Hand-off to the product experience
Every recommendation routes into the existing product page, keeping the conversation as a layer over the real catalog.

AI-powered support & operations
A centralized workspace for managing customer conversations, user and order details, tickets, and support workflows in one place. AI-assisted chat helps support teams understand context faster and respond more efficiently.

Outcome

Zenni demonstrates how an AI shopping agent can extend an existing e-commerce experience across discovery, decision-making, checkout and post-purchase support without replacing the traditional interface.

The strongest conversational commerce experience is not about making everything chat-based. Conversation works best for understanding intent and providing guidance, while structured UI works better for comparison, configuration and completing important actions.

What I learned

Conversational commerce works best as an additional path, not a replacement one. The hardest design work was not the chat UI — it was deciding which parts of a shopping journey genuinely improve in conversation and which are better left to traditional UI.