Autovision
Read more
The Challenge
Searching for right automotive parts takes a lot of time and energy. Additionally, understanding what the right part is for your vehicle requires extensive research and knowledge. How might we help the AutoZone customer find what they are looking for faster?
Solution
Utilizing the power of generative AI, AutoZone will enhance the product discovery process for users and offer trustworthy advice at scale.
Role
UX Designer
Industry
Automotive
Timeline
June 2024 - August 2024
Team
Solo project
Tools
Power point
Claude
Figma
I spent my summer at AutoZone working as a UX designer. For my final project, I proposed a new search functionality utilizing generative AI
Autovision
Two prototypes, a signed in and signed out version. Shows how a user can use AI in the search and discovery process.
See prototype
02. Design challenge
How might I help the AutoZone mobile app customer find the right product faster?
03. Target user
Who I am helping
Utilizing research conducted by the DXR team, AutoVision targets a specific power user. He is represented by the persona Ambitious Alonso. He is a 20 year old automotive enthusiast who is constantly fixing and enhancing his ride. In an emergency, his first stop is always AutoZone.
04. Design process
Cross functional collaboration
I talked to many different teams to gain an understanding of the problem space
1
UXR team
I initially scheduled meetings with our in-house UXR team to understand mobile app users and consumer behavior on e-commerce platforms, and how this differs from in-store behavior.
2
Marketing team
I wanted to get a clear understanding of the most important and top selling categories. This was important to understand because it would help me determine what product to focus on for my designs.
3
Store Autozoner
Supporting the stores is an important part of the aftermarket automotive industry. I interviewed Taylor, who has been working at the Germantown AutoZone store for 2 years. I wanted AutoVision to mimics the advice a store AutoZoner would give to a customer.
4
Generative AI team
I needed to get a clear understanding of the capabilities of implementing a large language model into our mobile app.
5
Mobile app team
Once I designed a mid-fidelity version of my idea, I talked to the mobile app team to gather feedback. A Senior UX designer was able to provide valuable feedback on how I can create a full end to end experience by thinking about the signed in user and the signed out user
6
Analytics team
An important part of releasing a new feature is understanding the impact it could potentially have with numbers. The analytics team helped me find data on the business benefit and also how much this would cost.
05. Gathering insights
Designing a prototype
After getting a clear picture of the needs and benefits, I began designing a high fidelity version of AutoVision. I started by creating a mind map to highlight all the key features I wanted to include in the prototype.
06. Competitive analysis
Use of AI in ecommerce
Etsy Gift Mode
Users take a quiz that asks them questions about the occasion, the gift recipient's relationship to them etc. Then, utilizing Open AI’s GPT- 4, Etsy Gift Mode generates tailored gift recommendations. Thereby addressing common gift buying challenge of not knowing what to buy.
AMZ Rufus
Amazon is one of AutoZone's biggest competitors in the e-commerce space. They launched Rufus which is an AI chat bot.
07. Creating a design system
Utilizing AutoZone's Global Style Guide
AutoZone has an extensive library of components and a robust style guide. I utilized both to help design AutoVision. Since I was proposing a net new experience, I also designed new components to that would fit with the product.
08. Full End to End Experience
Product features
Signed out users
Add vehicles
For signed out users, they can add their vehicle make and model to ensure the most accurate and trustworthly advice.
Signed in users
Entire flow
Finding the right car parts can be difficult and having the ability to access past searches is a time saver. In order to encourage more users to create an account, past searches is a signed in only feature.
09. Large language models
Thinking Through Technical Feasibility
When choosing a third party vendor, I had to consider three key factors, price, quality, and speed. In the end, I believe working with Google's AI studio will be the best choice.
A model's quality is measured by many things. Some of the most important are: perplexity (how well a probability model predicts a sample), human evaluation, how well it translates other languages.
Speed is measured by output tokens per second. In other words, it is how quickly the model can generate text.
10. Evaluation and presentation
Presenting my project
During the last week of my internship I presented my idea to three directors in the e-commerce department and my fellow peers at AutoZone.
11. Next steps
Retrospective
Providing customers with trustworthy advice has always been a core tenant of AutoZone. Before launching the product to customers, I would recommend having an internal team validate proof of concept and monitor the results to ensure accuracy. Generative AI can be extremely beneficial to retailers. The technology is only getting cheaper and cheaper and more and more accurate. I can't wait to see what the future will hold.
My time at AutoZone was an incredible learning experience. I had the opportunity to join a talented team and build upon established design foundations. The final project I worked on enabled me to demonstrate my ability to think critically about the company's needs and envision an effective solution. I am especially grateful to all the people and teams who helped make my summer unforgettable!
