Optimizing AI Conversations: A Case Study on Personalized Shopping Assistance Frameworks
This video is featured in the Designing with AI 2025 playlist.
Summary
The Challenge Users engaging with the AI shopping assistant often felt constrained by limited options, excessive follow-up questions, and a lack of personalization. These shortcomings led to user fatigue, misunderstandings, and a subpar shopping experience. Insights from user research (UXR) and transcripts revealed that users wanted more intuitive, human-like interactions that catered to their unique needs. The Solution A robust, adaptable framework was designed to transform AI conversations into sales-like consultations. By breaking user queries into three core components—use-case, constraints, and preferences—the framework enabled the bot to understand intent and deliver relevant, personalized results. Key enhancements included: Allowing users to skip questions and navigate freely. Providing contextual help for technical queries. Transitioning to open-ended interactions after gathering essential details to prevent over-questioning. Displaying diverse and curated results aligned with user preferences.
Key Insights
-
•
Traditional conversational AI severely limits visual bandwidth compared to traditional e-commerce interfaces, hindering effective product exploration.
-
•
The catalog exploration paradox highlights the tension between rich visual browsing and conversational limitations in AI assistants.
-
•
The expectation gap occurs when AI fails to understand cultural or contextual nuances, eroding user trust and satisfaction.
-
•
User queries vary widely in specificity; treating all queries the same is a missed opportunity in AI design.
-
•
Structuring user queries into use case, constraints, and preferences enables AI to perform intelligent reasoning like a skilled salesperson.
-
•
Carefully crafted prompt design, rather than large-scale fine tuning, can effectively guide LLM reasoning for better conversational AI.
-
•
Adaptive questioning avoids unnecessary or repetitive queries, improving user experience and reducing frustration.
-
•
Designing conversational AI as reasoning systems rather than linear scripted flows fundamentally improves interaction quality.
-
•
A multidisciplinary team combining design, data, and engineering perspectives was crucial to solving this AI conversation challenge.
-
•
Shifting from seeing AI as a task execution tool to a trusted, empathetic guide changes user perception and engagement positively.
Notable Quotes
"We are designing how the assistant is trying to reason, moving from scripting responses to orchestrating the AI's cognitive process."
"It's not just about fixing the chatbot, it's about bridging the gap between user expectations and actual experiences."
"LLMs don't just match keywords or labels. They infer meaning, extract subtle nuances, and understand intent behind words."
"The catalog exploration paradox shows how current conversational AI restricts browsing and comparison, creating uncertainty."
"Users frequently felt overwhelmed and misunderstood; this was a significant barrier to a positive shopping experience."
"What if we actively taught the model how to break down and analyze queries like a skilled salesperson would?"
"The AI assistant starts with reasoning and not results, engaging users in intelligent conversation to understand needs."
"Don't design for a response, design for reasoning. Teach AI to think critically, analyze information, and arrive at logical conclusions."
"The assistant asks open-ended questions that focus on what truly matters: performance, budget, and preferences."
"We saw a 2.5 times increase in active monthly users and a 1.5 times increase in purchases attributed to the AI assistant."
Or choose a question:
More Videos
"Design ops’ job is like jazz rules—it provides a framework so creativity can flow without worrying about inventing structure mid-process."
Jim KalbachJazz Improvisation as a Model for Team Collaboration
November 6, 2017
"Testing your book idea through workshops or talks gives a forcing function that helps prioritize content."
Louis RosenfeldCoffee with Lou: Should You Write a (UX) Book?
March 7, 2024
"Building influence has been such a challenging part of growing in my own career. It takes a lot of communication and understanding others’ goals."
Catt Small Micah Bennett Brian Carr Jessica HarlleeWhat's Next for ICs: Exploring Staff and Principal Designer Roles
February 22, 2024
"Less than a third of respondents feel like they work somewhere with a clear investment in customer-driven decision making."
Marieke McCloskeyUser Science: Product Analytics & User Research
March 11, 2021
"User recordings are your most valuable asset but have become riskier due to biometric privacy laws."
Llewyn Paine[Demo] Deploying AI doppelgangers to de-identify user research recordings
June 5, 2024
"Correlation doesn’t imply causation, but sometimes two things happening together do imply a causal mechanism."
Joshua NobleCasual Inference
October 6, 2023
"We’re biologically wired to respond to intellectual challenges the same way as physical threats."
Sara LogelYour Colleagues are Your Users Too
March 29, 2023
"Sponsor sessions may require registration by email to prevent Zoom bombings and protect session integrity."
Bria Alexander Louis RosenfeldWelcome
January 8, 2024
"If you have to learn a workaround, you want to learn it once and reuse it again and again."
Sam ProulxOnline Shopping: Designing an Accessible Experience
June 7, 2023
Latest Books All books
Dig deeper with the Rosenbot
How does healthcare UX balance creativity with strict regulatory compliance and safety requirements?
How can research repositories adapt to handle dated content and decide when to retire old insights?
How do research operations teams manage the tension between consolidating tools and needing specialized capabilities?