Summer 2024 Internship

OVERVIEW

Optimized customer product search experiences through an AI-driven, personalized digital product.

As the only intern for this project in the gen AI team, I aimed to help users discover relevant items more quickly and effortlessly through generative AI. Initially, the product lacked understanding of user context before, during, and after searches. I addressed this by designing empathetic personalization that anticipated user journeys and empowered decision-making with click-through rate data and gen AI.

Role

UXUI Design Intern

Company

adMarketplace

Location

New York, NY

Timeline

June - August 2024

Design

⁃ Journey map, Interaction Diagram, Defined the value propositions
⁃ Sketch, Wireframing, Low-fi and clickable prototyping in Figma
⁃ Final Deliverables
⁃ Built conversational chat experiences for consumers.

Research

⁃ Analysis and synthesis alpha group user test feedback after the demo.
⁃ User Interview and User Testing.
⁃ Defined a relationship between AI and customers to leverage Gen AI.

Collaboration

⁃ Key stakeholders in weekly cross-team meetings including AI/ML engineers, Front/Back-end developers, Product team.
⁃ Presented AI-driven UX implementation opportunities.
⁃ I worked closely with the design manager and the product manager.

Role

UXUI Design Intern

Company

adMarketplace

Location

New York, NY

Duration

June - Agust 2024

Design

⁃ Journey map, Interaction Diagram, Defined the value propositions
⁃ Sketch, Wireframing, Low-fidelity prototyping, Clickable prototyping in Figma, Final Deliverables
⁃ Built conversational chat experiences for consumers.

Research

⁃ Analysis and synthesis alpha group user test feedback after the demo.
⁃ User Interview and User Testing.
⁃ Defined a relationship between AI and customers to leverage Gen AI.

Collaboration

OVERVIEW

Optimizing customer product search experiences through an AI-driven personalized digital product.

Aimed to help users discover relevant items more quickly and effortlessly through generative AI. Initially, the product missed the specific context and needs of users before, during, and after searches. I addressed it by designing empathetic personalization to anticipate user journeys and empower their decision-making with click-through rate data and gen AI.

Role

UXUI Design Intern

Company

adMarketplace

Location

New York, NY

Duration

June 2024 - August 2024

Responsibilities

• Optimized customer product search experience by defining the relationship between AI and customers and leveraging Gen AI.

• Designed AI-powered personal shopping assistant digital product which included researching and building conversational chat experiences for users.

Design

• Journey map, interaction diagram

• Defined the value propositions

• Sketch, Wireframing, Low-fidelity prototyping, Clickable prototyping in Figma

• Final Deliverables

Research

• Analysis and synthesis alpha group user test feedback after the demo

• User Interview and User Testing

Collaboration

• Key stakeholders in weekly cross-team meetings (AI/ML engineers, Front/Back-end developers, Product team)

• Presented AI-driven UX implementation opportunities

I was the only intern responsible for this project in the gen AI team, and I worked closely with the design manager and the product manager.

Role

UXUI Design Intern

Research

Defined a relationship btn AI and humans

Implemented conversational experiences

Analyzed and synthesized alpha group user test feedback after the demo

Design

Built frameworks, including ecosystem map, journey map, and interaction diagram for the shopping journey

Defined the value propositions

Designed Interactions

Created Sketch, Wireframing, and low fi prototyping

Crafted clickable prototypes in Figma

Conducted onboarding process and customer profiles

Collaborated with other stakeholders

Attended Weekly meeting with AI, ML, frontend, and backend developers

Participated in weekly product team meetings

Presented AI-driven UX implementation opportunities for the product to the entire product team

Company name

adMarketplace

Employment type

Advertising Technology

Location

New York City Metropolitan Area, NY

Duration

June 2024 - August 2024 (12 weeks)

FINAL DESIGN

Scenario 1. A non-linear shopping journey where the AI understands user clicks and suggests products.

Personalized, context-based suggestions

Jobs To Be Done

Discover perfect shoe that matches my style in a large number of options.

Problem

The limited mobile screen only displays four options at a time, making it frustrating and time-consuming for customers.

How To Solve (use case)

Help users narrow down their choices and make easier purchase decisions by using AI to personalize suggested options and questions based on their browsing behavior.

Scenario 2. AI engages the user in a conversation by asking contextually relevant questions, helping them discover the right product through informed decision-making.

Research-based, conversational support

Jobs To Be Done

Find cosmetic products that make skin glow, suited to my skin type.

Problem

Exhausted from compiling information on required product features and confused about what is necessary and which details should be prioritized.

How To Solve (use case)

AI provides research-based questions and answer options to help users consider key product features before making a purchase—making the search journey easier and smarter.

Ask AI

Jobs To Be Done

Obtain immediate clarity from specific, self-focused questions.

Problem

Additional follow-up questions require users to leave the app to search on Google.

How To Solve (use case)

The Ask AI interface is consistently placed to allow access and utilization at any time.

Scenario 3. AI helps users find outfits by using event details and user preferences, creating a joyful, relaxed shopping process for important events like weddings or dates.

Joyful, mix-and-match search

Jobs To Be Done

Prepare a confience-boosting outfit for a special event.

Problem

Traditional, time-consuming searches add stress to the shopping experience.

How To Solve (use case)

AI generates various appealing combinations for consumers to engage in a mix-and-match activity. AI-driven searches de-stress the process and allow for an enjoyable search experience.

ITERATION

Optimized mobile screen interface to better accommodate AI Chat with product results.

"It's not easy to distinguish between chatting and product results when asking and answering questions to AI."
- User feedback

Modified the design principle of 'consistency' to fit the context of AI chat, encouraging users to respond more easily.

Before

After

What I did

Modified the design principle of 'consistency' to fit the context of AI chat, encouraging users to respond more easily.

Add background color to improve distinction between two different interactions: chatting and product results.

Divide the content into manageable chunks to make it less overwhelming and more intuitive.

For consistency, implement a fixed chat space rather than one that changes based on the amount of conversation.

Transition to full screen and hide the product results section to help users focus on their question interaction.

Optimized the accessibility of interactions with Ask AI floating button.

1
Improved clarity by consolidating scattered AI buttons into a single floating button.
‍

A major usability feature is the inclusion of follow-up questions.
‍

2
Conversations translate into an overlay interface, enabling AI-driven interactions and flexible product exploration.

A major usability feature is that users manually answer the AI's questions.
‍

3
Enabled users to resize the AI chat screen by dragging, depending on their contextual needs.
‍

When users want to ask follow-up questions, the chat interface expands to full screen.

CONSUMER SHOPPING JOURNEY MAP FOR PRODUCT SEARCH & BROWSING

Product-oriented search and general browsing may start differently, but follow the same four steps — each laborious and overwhelming to consumers.

Research about the item

Define requirements

Comparison

Evaluation

● Research ● Define requirements ● Comparison ● Evaluation

Product-oriented search and general browsing may start differently, but follow the same four steps—each laborious and overwhelming to users.
Consumers feel overwhelmed and burdened by the complexity of researching, defining, comparing, and evaluating product requirements.

An analysis of users shopping behavior patterns demonstrated that consumers demand flexible context- and situation-based assistance during their journey.

An analysis of users shopping behavior patterns demonstrated that there are five necessary steps which all shoppers must complete to find matched items such as general research, information gathering, product comparison, product evaluation, and narrowing down search results. Whether they are doing general browsing or a product-oriented search, users still must to through this process. As a result, users demand flexible context- and situation-based assistance during their journey.

INTERACTION DIAGRAM

As users navigate nonlinear journeys and shift interests easily, adaptive contextual support can guide them toward purchase.

A consumer’s shopping goal can change mid-journey and across multiple touchpoints. For example, the user starts the journey on Instagram where they are intrigued by an ad about black sneakers. The user clicks on the ad and scrolls down the page, where a black silk dress catches their eye. The consumer then clicks on the dress to learn more. It’s important to know that the goal may shift throughout the course of the shopping journey, like here when the user starts by shopping for other items or nothing instead ends up shopping for the sneakers. An analysis of consumers shopping behavior patterns demonstrated that consumers demand flexible context- and situation-based assistance during their journey.

A consumer’s shopping goal can shift mid-journey and across multiple touchpoints. For example, a user might start on Instagram interested in black sneakers, but end up exploring a black silk dress—or decide not to buy anything at all. This illustrates how goals can evolve throughout the journey. Behavioral analysis shows that consumers need flexible, context- and situation-aware assistance to support these shifts.

IMPLEMENT CONVERSATIONAL DESIGN

AI-personalized communication supports shopping journeys flexibly across multiple contexts.

By questioning “How do we design a relationship?”, I wanted to describe AI as a personal assistant which communicates with suggestions by sharing history and preferences while maintaining the initial search goal, which can shift depending on the user’s context. The focus is on understanding the context. Ask AI's aim is to collaborate and learn from interaction between AI and users to help them to discover ideal items. Gen AI provides context-based queries and generates product results based on the user's answers. AI researches product information that a user needs to consider before making a purchase decision.

By asking ‘How do we design a relationship?’, I explored AI as a personal assistant that adapts to user context and shifting goals. Ask AI aims to collaborate with users, using Gen AI to suggest products through context-aware queries and shared preferences, and researches product information that consumers need to consider before making a purchase decision.

When and how users meet Gen AI and what and how Gen AI helps the users. I created crucial questions by answering the curiosities regarding what users need. How does the solution meet the users’ needs? How does Gen AI meet user needs? How does Gen AI support decision making? And what kind of datasets are needed to improve Gen AI? AI generates questions to help users navigate and to check the current status to make wise purchase decisions. The user selects an answer from options generated by AI, based on research users need to consider before making a purchase decision.

I explored when and how users interact with Gen AI during chat, and how it supports decision-making. By identifying key user needs, I framed questions such as: How does Gen AI meet user goals? What datasets improve its performance? The AI generates contextual questions to guide users and assess their purchase readiness. Users select answers from AI-generated options based on relevant product research.

ANALYSIS & SYNTHESIS OF USER TEST FEEDBACK AFTER THE FIRST DEMO

Complementary relationships are built between users and AI.

“The more prompts I answered, the more specific product results I got”
- 18-24, Female, In the past week, 3/5

"Help me to recognize the important product features"
- 35-44, Male, In the past week, 4/5

The user is learning to navigate the AI

Narrowed down options based on my likes/dislikes - would be cool if the algorithm optimized towards my trending likes/dislikes.
‍
23-24, Female, In the past week, 4/5

I was impressed by how smart the prompts were and found them to be helpful in my search.

‍

18-24, Female, In the past week, 4/5

AI is learning the user

I usually started with a generic item like “shoes and it did an ok job at narrowing it down, but it would have been more efficient if my initial search was “men’s running shoe size 0”

23-24, Male, In the past 6 months, 3/5

Again, the prompt were generally good fine from a testing perspective. They were within the realm of what I was interested it in.

45-54, Male, In the past week, 3/5

There was mutual learning relationship—users are learning how the AI is working and the AI is adapting to the users input. Feedback from the product's Alpha Launch surveys show that users naturally learn the product features by recognizing how their queries influence product results, and the product improves future search results based on users' feedback on satisfaction and relevancy.

WHAT I LEARNED

Designed engagement-rich AI-driven interface to empower users’ product search experiences.

While designing gen AI-driven UX and UI, I aimed to support users in their contextualized journeys by offering more personalized solutions, which required real-time updates. However, this functionality is typically only feasible at large tech companies, so I had to find alternatives, such as anticipating user needs by analyzing click-through data history. To collect more data, I worked to encourage user engagement with AI by enhancing accessibility in the UI. If this limitation is addressed, I believe contextual AI support can be developed to flexibly accommodate users’ unexpected and evolving needs throughout their journeys.

FINAL DESIGN

Scenario 1. A non-linear shopping journey where the AI understands user clicks and suggests products.

Personalized, context-based suggestions

Jobs To Be Done

Discover perfect shoe that matches my style in a large number of options.

Problem

The limited mobile screen only displays four options at a time, making it frustrating and time-consuming for customers.

How To Solve (use case)

Help users narrow down their choices and make easier purchase decisions by using AI to personalize suggested options and questions based on their browsing behavior.

Scenario 2. AI engages the user in a conversation by asking contextually relevant questions, helping them discover the right product through informed decision-making.

Research-based, conversational support

Jobs To Be Done

Find cosmetic products that make skin glow, suited to my skin type.

Problem

Exhausted from compiling information on required product features and confused about what is necessary and which details should be prioritized.

How To Solve (use case)

AI provides research-based questions and options for answers to help users consider the required product features before making purchase decisions, making the search journey easier and smarter.

Ask AI

Jobs To Be Done

Obtain immediate clarity from specific, self-focused questions.

Problem

Additional follow-up questions require users to leave the app to search on Google.

How To Solve (use case)

The Ask AI interface is consistently placed to allow access and utilization at any time.

Scenario 3. AI helps users find outfits by using event details and user preferences, creating a joyful, relaxed shopping process for important events like weddings or dates.

Joyful, mix-and-match search

Jobs To Be Done

Prepare a confience-boosting outfit for a special event.

Problem

Traditional, time-consuming searches add stress to the shopping experience.

How To Solve (use case)

AI generates various appealing combinations for consumers to engage in a mix-and-match activity. AI-driven searches de-stress the process and allow for an enjoyable search experience.

3 KEY TAKEAWAYS (FEATURES/INSIGHT)

Context-based suggestions: Real-time observation by tracking user clicks.

Shopping journeys are not always linear, as revealed by analyzing the user's shopping behavior patterns while creating journey maps and interaction diagrams. Although general browsing and product-oriented searches may begin at different points, they share several common steps, such as general research, information gathering, product comparison, product evaluation, and narrowing down search results. Therefore, users need flexible assistance based on their situation.

Conversation support: Gen AI creates personalized and research-based queries.

Ask AI is designed aim to collaborate and learn each other AI and the users to help them to discover ideal items. Gen ai provides context based queries (questions) and product results based on the user's answer. AI researches a product information that a user need to consider before making purchase decision and narrowing down their selection.

Search with joy: Playing with their prefered items to find matched items.

The consumers feel pressure to prepare the big event. The entertainment help a user find the ideal items with joy and reduce cumbersome or pressure. For example, I need cute and lovely outfits for my first dating or I need a dark blue dress for my friend’s wedding in Chicago.