In-Car Shopping Agent Design
@Grok
Client:
Tesla
Deliverables:
User Research, UXUI
Year:
2026
01 CONTEXT
A. Physical Ecosystem
Tesla was building large charging hubs like Oasis and exploring services beyond charging.
B. User Signal
On X, Tesla owners were imagining new ways to access food while charging.
C. Technical Enabler
The engineering team had integrated Grok into vehicles, allowing users to activate it simply by saying “Hi, Grok.”
02 CHALLENGES
Enable shopping without demanding attention
Traditional shopping relies on browsing, comparison, and repeated decisions—behaviors that conflict with the limited attention available in a vehicle.
B. Navigating Ambiguity
Navigating high ambiguity in an exploratory project by leading the research, framing the core problem, and defining the solution from scratch.
03 dESIGN STRATEGY
Make Agentic AI Tangible
Think of grok as a waiter
To make an abstract agentic experience tangible, I started with a familiar mental model: ordering at a restaurant, with Grok acting as the waiter.
Scenario 01
First-time customer, high guidance
Grok can start with popular options, then progressively guide them through choices like combos, sides, drinks, and customization.
Scenario 02
Returning customer, high personalization
A returning customer may already have established preferences. Grok can recall past orders, surface familiar options, and ask whether they want to reorder or customize them.
B. Define the user journey
I translated research and user stories into an end-to-end journey, defining when Grok should anticipate, guide, confirm, and act.
04 Defining the Agent
Multimodal Fluency (VUI)
Keep interaction hands-free, using visuals only for lightweight confirmation.
[Customer: “Hi Grok, take me to a nearby Supercharger.”]
B. Context Awareness
Grok uses route and time context to proactively suggest relevant nearby options.
[Grok: I noticed that there are lots of amenities nearby and it’s near lunchtime, would you like to order lunch ahead?
Customer: Yes please.]
C. Guided decision-making
Turn open-ended decisions into a sequence of lightweight choices.First-time customer
First-time customer
[Grok: Would you like to have a combo, a burger or a milkshake?
Customer: A cheeseburger.
Grok: Do you want fries and drinks to come with the burger?
Customer: A Coke Zero.
Grok: Done. Ready to check out?
Customer: Yes.]
Returning customer
[Grok: Do you have something in mind that you’d like to order?
Customer: Yes I would like to order a Garlic Single Steakburger with no cheese just as usual.
Grok: Is that all for today?
Customer: Yes.]
D. Preserve User Control
Let Grok prepare actions, but require explicit confirmation for purchases and costs.
[Grok: It will be 16.61 dollars in total for 2 items. Do you want to use your credit card?
Customer: Yes.]
E. Agentic Follow-through
Continue supporting the user through navigation, order status, and pickup.
[Grok: Order placed! It may take 10-15 minutes. Let’s navigate to the Supercharger first then go get your food.
(After a short while)
Grok: Order ready! Pickup at #5.
05 BUSINESS OPPORTUNITY
Designed a scalable service layer that connects charging, mobility, and commerce across Tesla’s ecosystem.
0→1
Agentic commerce model
93.3%
Potential Supercharger site coverage
~$150B
U.S. QSR market opportunity