Boll & Branch's AI Virtual Bed Designer Doubles Average Order Value: A Cross-Selling Case That Moves In-Store Styling Advice Into AI
Luxury bedding brand Boll & Branch lets shoppers layer its bedding onto a photo of their own bed, in stores and online, and says users have doubled their average order value. We explain how it works, how to read the number, and what it means for AI-driven bundle recommendations.
Key Takeaways
- US luxury bedding brand Boll & Branch has been testing its first AI tool, the virtual bed designer, since April 2026. Shoppers layer bedding onto a photo of their own bed to try combinations. In stores, ambassadors use it alongside customers, and the brand is also testing it on its website
- Customers who use the tool have doubled their average order value, and demand has surfaced for throw blankets and decorative pillows, categories the brand did not sell much of. The notable move is using AI to deliver the know-how of an in-home bed styling service without entering anyone's bedroom
- The baseline and usage rate are undisclosed, and the tool may simply attract shoppers who were already ready to spend. For merchants, turning the "what goes with what" advice used in clienteling into data is preparation that pays off both in their own tools and with external AI agents
Boll & Branch Reports a Doubled Order Value From Its First AI Tool
Boll & Branch is using AI to scale its bed installation program.
www.retailbrew.comKatia Unlu, chief commercial officer of US luxury bedding brand Boll & Branch, shared results from the company's first AI tool during a panel at the industry conference eTail Boston. According to Retail Brew's report on September 24, 2026, customers who shop using the "virtual bed designer," which the brand began testing in April, have on average doubled their order value.
"It's quite substantial in terms of just the revenue," Unlu said, before pointing to a second effect. Boll & Branch is best known for its sheets, yet customers using the tool have driven up demand for decorative pieces such as throw blankets and pillows. The tool has surfaced categories the brand does not sell a ton of.
In stores, the brand's ambassadors (sales associates) use the tool alongside customers, and it is also being tested on the website. Its origin lies in in-store service.
It Started With a Service That Enters the Bedroom
Boll & Branch launched in 2014 as a DTC (direct-to-consumer) brand. Before opening its own stores, it tested the waters through partnerships with Nordstrom and Bloomingdale's, and only then moved to standalone locations. It opened its first store in Short Hills, New Jersey, in 2017, and according to a Retail Brew article from November 2025, it nearly doubled its footprint from eight to 15 stores over the preceding year. The latest report says it will soon have 19 stores.
Unlu describes stores as a "repeat and lifetime value accelerator." CEO Scott Tannen has also said the brand favors suburban markets where homes have more bedrooms. The stores are designed as bases for long relationships with individual customers.
Built on this store network is an in-home program called Room Service. After a customer buys bedding, stylists from the store visit the home, bring a steamer and make up the bed four-star-hotel style, down to candies on the pillow. Figures the brand shared in a Retail Brew article from March 2026 show that new customers who use the service have a repeat rate of 67%, well above the 46% for new customers who don't, and a lifetime value 188% higher.
The problem was scale. Because not everybody wants people in their bedrooms, Unlu said the program has been "hard to scale."
What is easy to overlook is that the value of Room Service is not only in making the bed. Stylists see the customer's actual bedroom and can suggest what to add to what is already there. The virtual bed designer can be read as an attempt to extract just the "suggest combinations while looking at the real bedroom" part and deliver it without entering the room.
Store strategies among DTC bedding brands are not uniformly successful, though. Rival Parachute announced in June 2025 that it would close 19 stores and shift its focus to partnerships with retailers such as Target.
How the Virtual Bed Designer Works
The official page keeps it simple. Shoppers upload a photo of their bed or choose one of the brand's, compare styles and colors of sheets and duvets layered onto it, and add the combination they like to the cart to check out. The page also lists several "custom bundle" examples that combine a comforter set, sheet set, bed blanket and decorative pillow cover.
The key is being able to swap products one at a time while checking how the whole bed looks. A customer testimonial on the page says that within seconds she saw a patterned sheet would look terrible in her room, then tried several solid colors and chose one.
Retail Brew's article does not name the technology provider, but the "Start Designing" button on the official page leads to a domain run by AI company Hemsy. Hemsy lists Boll & Branch as a customer on its website and features a comment from Julia Schneider, Boll & Branch's VP of Commerce, who praises its accurate rendering and an intuitive interface that bridges browsing and buying.
According to Hemsy's developer documentation, its rendering is not image generation from scratch. It edits only the product slots in the user's real photo, keeping the room, camera angle and lighting intact, while product images and color swatches lock silhouette, material and colorway. The documentation contrasts this with asking a model to invent a bedroom with a navy duvet. The claim is that image editing constrained by the real catalog is what makes the output usable for purchase decisions.
Other Hemsy features overlap with what Boll & Branch describes.
- Four launch modes: a guided flow that walks shoppers step by step to a complete look, a bundle mode that loads a predefined set at once, a sandbox where shoppers build a look piece by piece from the catalog, and a mode launched from the product page
- A "Stylist Mode" for store associates and customer service teams, which Hemsy describes as a surface built for the sales floor and chat upsell
- Analytics showing visualization volume and attributed revenue by surface
The link on Boll & Branch's official page includes a parameter that launches a guided flow. Whether store ambassadors use Stylist Mode, and which analytics the brand uses, is undisclosed. Hemsy's website cites a "2x lift to AOV" as a merchant result, but does not say which merchant or period the figure covers.
How to Read "Double the Order Value"
The doubling is eye-catching, but it is the only figure disclosed. It is not clear what it is compared against, whether customers who did not use the tool or the same customer segment before the launch. The usage rate, number of users, split between stores and web, contribution to total revenue and return rate are all undisclosed.
Selection bias also matters. Shoppers who take the trouble to photograph their bed and combine bedding are likely already planning to replace a full bed setup, and might have placed large orders without the tool. In stores, it is also hard to separate the effect of an ambassador making suggestions alongside the customer from the effect of the AI rendering. The Room Service repeat and lifetime value figures share the same structure, and it remains possible that already engaged customers choose the service.
The observation that decorative categories surfaced carries a different kind of meaning, regardless of the size of the number. Showing the bed as a finished look reveals gaps that sheets alone cannot fill, and that creates room for suggestions.
Home goods have seen a number of tools that show a finished look from a photo.
| Aspect | Boll & Branch Virtual Bed Designer | IKEA Kreativ | Wayfair Muse |
|---|---|---|---|
| Launch | Testing began April 2026 | June 2022 (US) | February 2025 (replaced Decorify) |
| What the shopper provides | A photo of their own bed, or one of the brand's sample photos | Several photos of a room, assembled into a 3D replica | A trending style, a text description, or a photo of a room |
| Unit of the proposal | The bedding combination for one bed | Furniture layout for a whole room | A whole-room style plus product ideas by category |
| Human involvement | In stores, ambassadors use it alongside the customer | Self-service (with placement assistance) | Self-service |
| Published results | Users' average order value doubled (baseline undisclosed) | Not found in the sources reviewed for this article | Undisclosed (success to be judged by engagement and repeat use) |
IKEA Kreativ is an elaborate system that reconstructs an entire room in 3D for placing furniture, and Wayfair replaced Decorify, the photo-based tool it launched in 2023, with Muse in 2025. According to Home Accents Today, a Wayfair executive said engagement and repeat usage would be the metrics to track for Muse, and no revenue contribution figures were given. By comparison, Boll & Branch narrows its tool to a single bed and frames each proposal as "the bedding combination to add to the cart today." The narrow scope makes both rendering accuracy and the distance to purchase easier to control.
Where It Meets Agentic Commerce: Building a Product Set From a Goal
This tool is not an AI agent that completes purchases on the customer's behalf. The customer chooses the combination and decides to buy. Even so, it offers lessons for agentic commerce, in which AI agents handle or assist everything from product discovery to purchase.
The shopper's starting point is a goal, "I want my bedroom to look like this," not a specific product name. Assembling sheets, a duvet, a blanket and pillows into a single answer from that goal has the same shape as what people expect when they ask a conversational AI for "a calm-looking bedroom." Rather than searching for and comparing single items, assembling and proposing a set of products (a bundle) from a goal is becoming a core function of AI-mediated shopping.
Hemsy is moving in this direction as well. It offers a rendering API for AI agents and a server for MCP (Model Context Protocol), the standard through which agents call external tools. Its documentation gives the example of a shopping agent recommending a duvet cover and a user wanting to see it on their own bed. If the product belongs to a Hemsy merchant, the color swatches and placement rules that merchant configured for its own storefront are applied automatically, improving render fidelity. Catalog information a brand prepares for its own tool thus also shapes the quality of suggestions made through external AI agents.
On the scale of AI-mediated shopping, Salesforce estimates that AI and agents influenced 20% of retail sales, or $262 billion, during the 2025 holiday season (November 1 to December 31). It also says shoppers referred to retailer sites from AI search channels such as ChatGPT converted nine times more often than those from social media. These are, however, figures from a company that sells AI agent products, drawn from its own platform data, with a broad definition of "influence." The Boll & Branch case is likewise not a record of an AI agent buying on a customer's behalf, but an example of goal-driven combination suggestions lifting order value.
What Merchants Can Take From This Tomorrow
The easiest place to start is recording the combination suggestions made in stores and in chat. If you capture which products are recommended together and why, in terms of color, season, material and the feel of the room, that knowledge can be reused both in your own tools and in product data for AI agents. Just as Hemsy defines a bundle as a list of product IDs down to the color and size variant, combinations become reusable only once they are machine-readable.
When designing cross-sells, it is worth looking at the smaller items around your core products. At Boll & Branch, the demand that surfaced was for the blankets and pillows next to the sheets. Products that feel minor on a single product page make sense once placed within a finished look. Identifying which items in your assortment "sell best as part of a set" changes how you design suggestions.
Measurement needs to be settled before launch. Comparing tool users with non-users may only show that the tool selects shoppers who were already ready to buy. Only with designs such as comparing shoppers who viewed the same product page or visited the same store, separating sessions with a stylist present from those used alone, and tracking return rates can you explain what is inside a "2x."
Conclusion
Boll & Branch's virtual bed designer is an attempt to replace the advice of a stylist who enters the bedroom with a photo and AI rendering. The doubled order value and the surfacing of decorative categories suggest that proposals assembling a product set from a goal can move revenue, but neither the baseline nor the usage rate is visible yet. The next things to watch are whether the web test moves to a full rollout, and whether the same catalog information comes to be used in suggestions made through external AI agents.
