Ford AI Shopping Assistant
With AI shaping the future of how we shop, Ford wanted to create an advanced shopping assistant to help our users find parts and accessories for their vehicles quicker and with less room for error.
Challenge
With newer AI technology, weaker product descriptions, and users not always knowing what part or accessory they may need, we had to tackle many challenges simultaneously to drive success. I had to design an interface that was easy to use and could get the user to their desired item quickly. In addition, I had to train the AI to be more conversational, use plain language, and ask important vehicle questions without bombarding the user. When it came to product description and comparison, I crafted created backend reasoning for how we show case items with weaker descriptions while still keeping our users informed.
User Research
With a new technology, we wanted to understand more about what users wanted to know accomplish with their AI shopping assistant.
Saving Time
Users will often spend hours online reading reviews, finding deals, and figuring out what they need for their vehicle. With an AI assistant, they expect all these answers to come within a few prompts. They want to be able to order their product quickly and get back to their daily activities.
Accuracy
There are so many vehicles, parts, and accessories on the market that the chances of a user ordering the wrong item is high. Users want their assistant to direct them to the correct item the first time so they are not having to figure out which items do or do not work with their vehicle.
Analysis
Users across the board had the same needs; wanting to find items for their vehicles quickly and efficiently without having to do the heavy comparison lift themselves. They want their assistant to be friendly but also informative. They also don't want to waste time at the dealership with the wrong part being delivered and having to restart the order process. We took these learnings and crafted an experience that will empower our users to select their exact item for their vehicles without taking too much time from their day.
Product Screens
I weighed on different options for displaying products and how to narrow down options and ultimately landed on a side bar user experience.
Landing Page
Product Carousel
Add to Cart
Follow Up Questions
Ford utilizes sidebar activity for multiple portions of the user journey. To avoid any awkward overlays or misses in vehicle fitment, I brought some of these elements into the chat widow to streamline the process.
Vehicle Select
Dealer Select
Vehicle Select in Chat
Dealer Select in Chat
Training the Model
That's right, you're looking at (one of the) voices of Ford's shopping assistant. With artificial intelligence comes artificial personality. Working alongside Ford's brand and content teams, I crafted a rubric for how we wanted our assistant to sound. It needs to be friendly and straightforward while also knowing the users next move. The rubric also judges responses on a 1 to 4 scale on how helpful and conversational the responses are, aiding in the training of the conversation model.
​
One instance of training our model includes how we can showcase products with detailed descriptions and how to display weaker descriptions. With detailed descriptions we want to highlight reviews and listed uses. With weaker descriptions we wanted to utilize other elements such as price and shipment avalibilty.
Strong Product Descriptions
Weak Product Descriptions
Outcomes
Currently the AI Chat Assistant is in development. Based on some initial market research we are expecting to solve several key issues in the customer journey while bringing in profit.
10%
30%
Increase in revenue
Decrease in returns