Enhancing Associate Learning with AI.
Understanding our users mental model to create a personalized learning experience
-
The Team
Project Manager
Product Manager
8 Software Engineers
Mobile Learning Partners
-
My Role
UX Researcher and UX Designer
-
Timeline
November 2024 - March 2025
Background
PocketGuide is a mobile microlearning platform used by store associates across the U.S. and Canada. It supports in-aisle learning through challenges and reference content, helping associates build product knowledge and selling skills. Despite its success (18M completions in FY23), associates still struggled to answer customer FAQs in real time, often resorting to personal devices and inconsistent sources like Google or YouTube.
Discovery Research
To better understand this gap, we conducted in-person interviews and usability testing with store associates and leaders at pilot locations. Our goals were to:
Understand current usage patterns and learning behaviors in PocketGuide
Identify pain points and gaps in associate knowledge
Evaluate the appetite for AI tools like chatbots and search bars
Assess trust and usability of AI-generated responses
We used a structured interview script that included demographic questions, scenario-based tasks, and sliding scale metrics to evaluate confidence, trust, and usability. We observed how they searched, what tools they used (chatbot vs. search bar), and how they rated the relevance and accuracy of the responses.
The Problem
Our research confirmed that associates often feel unprepared to answer customer questions about product selection, usage, and compatibility. They described PocketGuide as overwhelming, with too many challenges that didn’t always align with their roles or learning goals. Many associates defaulted to external sources or asked peers for help, leading to inconsistent information and potential loss in sales.
Hypothesis
We believed that integrating AI-powered tools—specifically a chatbot and a search summary feature—into PocketGuide would:
Help associates find accurate answers faster
Increase associate confidence
Improve customer satisfaction and sales
Personalize learning by surfacing relevant content based on user queries
Research and Methodology
We piloted two AI features:
AI Chatbot ("Ask the Expert") at local Home Depot store (Nov 7–20, 2024)
Methods Used:
User interviews (29 associates, 4 store leaders)
Usability testing with scenario-based tasks
Contextual inquiry
Usage data analysis
AI Search Summary at local Home Depot store (Dec 12, 2024 – Feb 9, 2025)
I synthesized our user interviews by completing an affinity map and identifying key trends and insights.
Key Insights
-
Anticipatory Learning
PocketGuide is more effective as a foundational learning tool rather than a real-time support tool. It supports anticipatory learning that builds associate knowledge before they need it.
-
Integration Recommendation
Replace “Ask the Expert” with a universal AI assistant, like copilot. Also, introduce AI role-play scenarios to support more engaging, hands-on learning and mentorship.
-
AI Interaction Preferences
Associates prefer AI responses that are:
Concise and bulleted
Supported by images for product comparison
Prompted with related questions
Capable of scanning SKUs or identifying objects via photos
-
Trust and Caution
Most associates trust AI responses without question, assuming they are accurate. This highlights the need for:
Clear disclaimers about AI limitations
Education on the potential for hallucinated or inaccurate information
Usage Data
Add in graphs from presentation
AI Recommendations
AI Assistant
Methods Used:
User interviews (29 associates, 4 store leaders)
Usability testing with scenario-based tasks
Contextual inquiry
Usage data analysis
AI Role Play
Design Solutions
We believe that if we provide users with an accessible and streamlined communication experience that allows for personalized preferences, customers will be more engaged.
We know this to be true if we see increased traffic on Home Depot’s website and reduced time spent connecting to customer service.
We began the design phase by designating roles and responsibilities for designing wireframes and prototypes.
-
Laura Kluball and Ellie Park
Design conversations and voiceflow prototypes for Dan (Laura) and Dambi (Ellie)
-
Tammie Davis and Dana Seaman
Design wireframes and Figma prototypes for personalized mobile website profile
I learned how to use Voiceflow prototyping software by watching tutorial videos!
Video Source: Intro to Voiceflow | Capture, Random and Exit Blocks
Below is the annotated Voiceflow diagram that I created for “Do-It-For-Me Dan.”
Usability Testing
To ensure that our users were satisfied with the changes we made, we conducted usability testing with our grayscale Figma wireframes and our first iterations of the Voiceflow prototypes.
Mobile website usability test tasks:
Create and verify an account with the Home Depot.
Set up communication preferences in portal.
Sign into existing Home Depot account and make an audio call with closed captions.
Voiceflow usability test tasks:
Script: “You are frustrated because your battery-powered lawn mower that you just purchased 30 days ago is no longer working and you want to take it back and get a new one. The mower itself is fine, but the battery will no longer charge, even though the charging station is working. You have the receipt for the purchase. Make an audio caption call to ask for instructions about returning your lawn mower battery (not the mower or charging station itself) and follow the prompts, to accomplish your task.”
Script: “Yesterday, you ordered the Ryobi ONE Power Drill and Ryobi ONE 18 Volt Cordless 3/8 inches Drill/Driver Kit on the Home Depot website to pick up in the store. So you are calling Home Depot customer service to check the status of your order. Your zip code is 1 2 3 4 5.”
Insights and Changes Made Based on Usability Testing
-
Insight #1: Mobile Website
8/8 users were confused where to click when signing in or creating an account.
Our testing yielded the results that led to a sign-in button here. A clear call-to-action button.
-
Insight #2: Mobile Website
We also received a lot of feedback for an adjustment in the wording for our secondary call-to-action, changing “Skip for Now” to “Skip and Add Later.”
“I want the ‘Skip for Now’ button to say ‘Skip and Add Later,’ so I know that if I don’t set my language preferences now, I will be able to do it later."
- User Quote
-
Insight #3: Mobile Website
2/2 users found difficulty signing into account and making an audio closed captioning phone call.
“I like the chat banner. I wish I could contact someone from there.”
- User Quote
This led to making the “Chat with Us” banner available on every page, and allowing users to communicate in more ways than just the chat bot.
-
Insight #4: Voiceflow
Users were frustrated that words were not interchangeable in the phone system. I.e. “exchange” for “return.”
The 2nd iteration implemented an enhanced “capture feature” with the word exchange encoded in the flow. IVR systems will work best when AI software is implemented to capture the “big picture” of the user experience and we are able to continuously improve the algorithms that build these systems.
“I wanted it [IVR] to pick up on the word ‘exchange.’ It had ‘return,’ but I wanted to ‘exchange,’ not return… I liked that it recognized my order and asked if that was the one I was talking about.”
- User Quote
Prototypes AFTER Usability Testing
Voiceflow Prototype
Dan’s IVR experience AFTER usability testing.
Voiceflow Prototype
Dambi’s IVR experience AFTER usability testing.
Mobile Website Prototype
Reflection
The Home Depot project was challenging, but also incredibly rewarding. This was the first time that we were exposed to interactive voice response system design, so we had a lot to learn in a very short period of time. Throughout this three-week design sprint, we worked hard to meet with as many stakeholders as possible, so that we could learn about the project goals and objectives from all perspectives.
As a conversational designer, I watched countless tutorial videos and collaborated with my colleague, Ellie, to learn how to use Voiceflow prototyping software and problem solve when we ran into roadblocks.
If we had another 6 months to work on this project, we have so many ideas we would love to help implement. We feel that we could enhance the IVR and chatbot personality by giving it a voice that represents the mission and eight core values of The Home Depot. Additionally, we would continue to improve the accessibility of both the IVR system, and the website platforms. In the last section, next steps are listed in greater detail.
Next Steps
-
Interactive Voice Response System
Enhance personality
Increased context comprehension with AI
Additional guidance and prompting
Natural language
Empathetic response-match tone, language, need of customer
Increased accessibility
-
Website
Information architecture
Continued Personalization
ADA compliance in color choice
Search feature specific to tools (DIY projects)
Pinterest on profile
Autosave preferences
Language preferences change entire experience on website