Cached at:
10/21/26, 01:37 PM
**TL;DR:** OpenAI showcased a shopping assistant demo: in the fictional home furnishings app Vasper and Alder, users can get room styling suggestions, generate visualizations, adjust items to fit a budget, add products to their cart, and change their shipping address through natural language conversations, without manually navigating the interface.
## Setting: The Vasper and Alder Shopping Assistant
The demo begins by logging into the fictional home furnishings app Vasper and Alder. The user plans to redecorate a room before hosting friends for the holidays, so they start a conversation with the app’s shopping assistant, asking for inspiration to make a corner of the room cozier—specifically, to create a reading nook.
## Step 1: Describe the Need in Natural Language, and AI Suggests a Setup
The user asks: “Can you give me some inspiration for making this corner of the room cozier? I want to create a reading nook.”
The assistant responds in two parts:
1. **Written styling suggestions**: It suggests uploading a photo to help coordinate the furnishings and recommends specific items—a cream-colored chair, a light oak table, and a floor lamp with a warm linen shade.
2. **Generated visualization**: The assistant also starts generating a visualization to show the user what the finished reading nook would look like.
This demonstrates two core capabilities of the shopping assistant: understanding a loosely defined, contextual need (“a cozy reading nook”) and translating it into a specific combination of products, then using a visualization to help the user confirm their choice.
## Step 2: Set a Budget, and the Assistant Automatically Replaces Items
After viewing the visualization, the user says, “This looks good,” and adds a constraint: “Try to keep it under 250.”
The assistant responds by:
- Replacing the chair with a **lummy cream-colored chair**, keeping the entire setup **under 250 before tax**
- **Updating the visualization at the same time**, so the user can see how the budget-adjusted setup would look
The user responds, “Wow, that’s great. The visualization really helps,” and then moves on to checkout.
This step demonstrates the assistant’s ability to maintain constraints, such as a budget, across multiple conversational turns and update the visual output accordingly.
## Step 3: Proactive Recommendations at Checkout
Once checkout begins, the assistant does more than simply process payment. It proactively suggests an additional purchase:
> “One last thing: this natural striped throw is only $17.50 and would add a little warmth. Would you like to add it?”
The user finds the price reasonable and replies, “Add it to the cart,” and the assistant does so.
## Step 4: Delivery Reminder and Address Change
After adding the item, the assistant displays the product list and proactively flags an important delivery detail:
> “Pickup at the Berkeley store will take 5 days.”
Only then does the user realize they accidentally selected delivery to the Berkeley store when they intended to have the order sent home, so they ask to change it.
The assistant’s process:
1. **Ask for the new address**: “What is your new street address, city, state, and ZIP code?”
2. The user provides: **1515 3rd Street, San Francisco, California, 94158**
3. **Display a confirmation interface**: “Please confirm the address here when you’re ready.”
4. Once the user confirms that everything is correct, the assistant updates the order, with **delivery ultimately changed to October 9**
5. Closing message: “Thanks for your purchase, Alex.”
## Capabilities Demonstrated Throughout the Process
The demo covers the key stages of a complete shopping journey:
| Stage | Assistant’s Behavior |
| --- | --- |
| Exploring ideas | Understands contextual needs and suggests a specific combination of products |
| Visualization | Generates visualizations of the setup in real time |
| Budget constraints | Automatically replaces items to stay within a $250 limit and updates the visualization |
| Proactive recommendations | Suggests adding a $17.50 natural striped throw at checkout |
| Delivery reminders | Proactively notes that pickup at the Berkeley store takes 5 days |
| Order changes | Collects and confirms a new address, then updates the delivery date |
Throughout the process, the user never has to leave the conversation to manually filter products, compare prices, or modify the order—all operations are handled through natural language, while the assistant carries out actions such as generating visualizations, updating the cart, and changing delivery details. This is the “Shopping Agent” experience OpenAI demonstrated: turning the conversation itself into the shopping interface.
Source: Build a Better Shopping Experience with OpenAI | Shopping Agent Demo (https://www.youtube.com/watch?v=8n-SXxRIrm8)