Nearly a year after raising a mega seed round of $50 million, e-commerce veteran Julie Bornstein’s startup Daydream is releasing its AI-powered chatbot for shopping with a focus on fashion.
After testing the product with select users, the company on Wednesday is releasing the chatbot to all users in a public beta. People can sign up for the chatbot, which will ask them their name, birthdate, price range they shop in, and brand preferences, if any.
You can type a query like “I want a dress to wear to the wedding this summer in Paris” or upload an image and add more context through text to search for a clothing item.
Users can save any item to a collection they create or refine their search by typing in the chatbot on the left-hand side. If they like an item but want to modify some aspects, like colors or style, they can tap on the “Say More” button visible on any of the items to edit the search.
Based on the parameters users provided during onboarding and feedback the app gets from them, it saves different items. Daydream creates a style passport for them that drives a lot of suggestions. The web app also shows you daily inspiration for items and accessories that possibly match your taste.
Right now, Daydream doesn’t have an integrated checkout flow, so when a user clicks on an item, they are redirected to a merchant’s website to complete a transaction. The startup is taking a percentage cut out of each sale. Daydream said that at launch, it has more than 8,000 brands on the platform, with the company onboarding new merchants free of cost.
Bornstein, who has held executive positions at companies like Nordstrom, Urban Outfitters, Sephora, and Stitch Fix, said that in the past year, the company was working on technology to bring a catalog of different brands to one place and revamp search to suit AI.
“Having worked in e-commerce my whole career, search has always been the forgotten child, and it never worked very well. In some ways, people were trained to just be very narrow in the way that they search for anything in the fashion world. And even with my prior startup, we couldn’t get people to go beyond something like ‘a red dress, ‘” Bornstein told TechCrunch over a call.
“But once Chat GPT launched, you know, consumers started to get trained on how to think about the potential of prompting. And so what we’re trying to do is help allow you to ask whatever it is you wanna ask, whether it’s I have an occasion, or I have this need, or I’m looking for this very specific kind of thing,” she said.
Maria Belousova, who joined the company this year as CTO, said that Daydream did a lot of work on understanding the nuances of items in the catalog. She said that traditional search only showed shoppers items based on tags matching their keywords, which doesn’t work in today’s world, where customers are asking longer queries.
“We are doing quite a bit to understand the details of the product, such as knowing stylistic attributes like embellishments, silhouette, or even social attributes like who would wear this dress, such as a bride or a guest in a wedding. We are also using image visual recognition to satisfy detailed queries where a customer describes the exact product they want,” Belousova said.
Over the next year, Daydream will allow users to give the tool more explicit feedback, like “Don’t show me any four-inch heels.” It also plans to experiment with a feature that will allow users to ask for a good match with an existing item with personalized suggestions. Plus, it wants to lean into the aspect of social sharing by letting users share their saved items with friends and family for suggestions to buy. Another feature that Daydream is thinking about is to take an existing collection of another user and modify it for their own needs using AI.
While Daydream’s team has years of e-commerce experience and focuses on fashion, startups like Deft and Cherry are also building multimodal search for shopping. Meanwhile, tech giants like Amazon and Google are focusing on features that use AI to search multiple sites and find the right item for users.
