overview
AI Meal Assistant
Enable shoppers to make quicker meal decisions and add items to cart seamlessly by redesigning into an intent-driven, integrated tool within key shopping touchpoints.
I led the redesign of AI Meal Assistant from a vague third-party chatbot provided by vendor into a structured, embedded experience that seamlessly supported meal planning and carting. By anchoring the tool within key discovery surfaces and shifting to intent-driven UX, we increased conversion, reduced friction, and positioned AI as a true commerce enabler.
Duration: Duration of 9 weeks
Role: Lead Product Designer
Team: 1 Design Manager, 1 Product Manager, Dev Team, Cooklist AI Team
Outcome: Session Starts +22%, Add-to-Cart + 9.5%, Revenue + 3.5%
Design Showcase
Contextual AI meal builder journey that facilitates user intention and add-to-cart conversions.
Before
The original assistant lived in an isolated chat panel, leading to low visibility, unclear intent, and poor conversion from interaction to checkout.
AFTER
A contextual widget embedded across exploration pages to align the assistant with natural moments of user intent.
Pre-loaded prompt suggestions helped users bypass blank-state friction and quickly generate relevant meal ideas.
A streamlined results flow allowed users to review, edit, and confidently add entire meals to cart—transforming inspiration into action.
Problem
Low Sessions Start, High Abandonment, and No Trust Path to checkout.
Two major business indicators were significant lower than our expectations in the e-commerce funnel: High Abandonment: 60% of users dropped off before carting any items, Low Add-to-cart rate, Minimal Revenue Impact: Assistant-attributed purchases accounted for less than 1% of total digital sales.
Approaches
Starting from user intention, Rethinking User Flow, Initiating Collaboration.
To solve the above low in add-to-cart rate, I was tasked with the UX initiative to find out the root cause of the problem, and navigate through this ambiguity. I started off by talking to Wegmans Insiders to gather qualitative feedback to fill in the gaps of the metric numbers. Why they were not using? What went wrong? Qualitative feedback revealed that users felt unsure about what to expect about this chatbot, hard to find accurate results, and no way to add the result to cart in the chatbot which caused high abandonment.
How might we evolve the AI assistant from a vague and purely converesational chat interface into a high-conversion meal tool? I rebuilt the system from both UX and logic perspectives. Users mentioned that the major thing they want to get out of this chat was to build a meal and add ingredients to cart. So I spearheaded the user flow discussion for Meal Builder Widget with the Product team and discussed a prompt guidance syntax collaborating with the AI team.
Solutions
Intent-First, Flow-Backed, Embedded by Design
We sunset the generic chat entry point and introduced a Meal Builder widget embedded across recipe, department, and discovery pages—surfaced exactly when users were already exploring what to eat. Key design changes included: Predefined prompts and predictive tags to eliminate blank-state hesitation, A structured meal-building experience that let users customize entrées, sides, and dietary preferences, A carting flow with confirmation, edit, and “finishing touches” upsells to drive confidence and value. Underneath, we mapped user input to a refined GPT syntax logic, optimized through collaboration with Cooklist, and built scalable pathways to grocery inventory and fulfillment.
Impact
From Tool to Funnel: Measurable Gains in Adoption and Revenue
Session Starts
Add-to-cart conversion rate
Assistant-attributed Revenue
The redesign not only improved UX clarity but proved that intelligent systems must be designed around behavioral pathways—not just technical capabilities.
Due to NDA, I can only show so much here. Would love to chat and share more details on the design thinking journey.