KODIF AI Shopper - Designing Conversational Ecommerce
Project Type
AI Chatbot UX Redesign
Project Timeline
Oct 2025 – Dec 2025
Context & Problem
Where the Chatbot Fell Short and Why It Mattered
KODIF's chatbot was built for post-purchase support - order status, returns, and FAQs but wasn't designed to help users decide what to buy. As KODIF expanded into ecommerce, users frequently dropped off when they wanted recommendations or comparisons. KODIF's chatbot handled 50K+ monthly conversations across retail clients, but 40% of users abandoned when asking comparison questions—representing significant lost revenue for clients paying per-conversation. When conversations reached pre-purchase decision points, text-only responses broke trust. Users left the chat to browse the website, fragmenting the journey and increasing abandonment.
Early wireframes showing the existing text-only support experience
The Challenge: Transform a support tool into a decision engine without breaking what already worked. Using Nike's catalog as a stress test helped us design for enterprise scale, not just demo conditions.
Research & Constraints
I interviewed 6 CX leaders and 3 ecommerce managers at KODIF clients, analyzed 200+ support transcripts, and tested prototypes with 10 end shoppers. Focus was on buyer needs (CX teams) balanced against user behavior (shoppers).
Key insights:
Technical constraints:
These constraints were validated in collaboration with engineering and CX teams, shaping early decisions around structured inputs, fallback states, and catalog handling.
Design Strategy
From Reactive Answers to Guided Decisions
I structured the solution around three interconnected flows, each aligned to a key decision moment in the shopping journey. The goal wasn’t automation - it was knowing when AI should guide the user and when the interface should take over.
Scalability Considerations:
Guided Discovery
Structured inputs (keyword/category/image) replaced open-ended prompts, reducing ambiguous queries by 60% in testing. Visual product cards surface immediately - no scrolling through generated text.
Tradeoff: Voice deprioritized in V1 due to 30% lower intent accuracy vs. structured input.
Smart Comparison
Side-by-side view for up to 4 products using structured attributes (size, price, reviews), not AI-generated prose.
Key Decision: Capped at 4 based on mobile testing—5+ items caused users to restart from scratch due to cognitive overload and lost context on smaller screens.
This directly addressed the #1 abandonment trigger from research.
Checkout & Unified Support
Checkout happens inside the same conversation, creating a continuous experience from discovery through post-purchase support.
Unified checkout inside chat: review cart → confirm address → pay → done. Post-purchase support continues in the same thread. One conversation, end-to-end.
Evidence from Early Testing
Tested with 10 participants in moderated sessions + 12 unmoderated tasks.
Visual cards cut time-to-product-selection by 50% vs. text-only (avg 45s → 22s)
9 out of 10 participants completed "compare 3 items and checkout" without leaving chat
6/6 CX leaders rated comparison flow as "critical missing capability"- validating prioritization over personalization or voice
Directional, not production data but strong enough to shift KODIF's ecommerce roadmap. This framework now underpins conversations with enterprise retailers evaluating KODIF.
What This Enabled for KODIF
Beyond the UX improvements, this work repositioned KODIF strategically. The company went from "post-purchase support tool" to "full-funnel commerce platform" in prospect conversations. The framework now anchors KODIF's ecommerce roadmap and has been presented to 3 major retail prospects in Q4 2025 sales cycles. Unshipped, but validated as strategically critical by customer teams and leadership.