Julie Bornstein thought It would be a cinch to implement his vision for an AI startup. His Résumé in Digital Commerce includes IndecCable: VP of Ecommerce at Nordstrom, COO of the Quarst Stitch Fix, and the personalization platform planet acquired by Pinterest. Fashion has been its claim ever since it became the inhaling high of the Syracuse boat spreading at seventeen and hanging in the local malls. So he felt it was a good idea to build a company for customers to find the perfect clothes using AI.
The truth was more difficult than he expected. I had breakfast recently with Bonystein and Maria Belousova, to learn about his startup, broke, funded by $50 million from VCs from VCS from Google like Google stuff. This conversation took an unexpected turn as the women delved into the surprising difficulty of translating the magic of AI systems into something people find useful.t
His story helps explain something. My first newsletter announced that 2025 would be the year of the AI app. Although there are many such apps, they haven’t changed the world as much as I expected. Since chatgpt started in late 2022, people have been blown away by the tricks of AI, but study after study has shown that the technology has not yet made a huge impact on productivity. (One: coding.) A study published in August found that 19 Ayi Enterprise Pilot projects did not deliver the estimated value. I think the productivity increase is close, but it is taking longer than people expect. Listening to start-up stories like a daydream that forces a break-up gives hope that persistence and perseverance really can be about making those things happen.
Fashionista fails
Bornststein’s Original Pitch to VCS seemed obvious: Use AI to solve tricky fashion problems by matching customers with perfect clothes, which they’ll be happy to pay for. (Dudream will take a cut.) You’d think the setup would be easy – just connect to a model API like chatgpt and you’re good to go, right? Um, no. Signing up to over 265 partners, with access to over 2 million products from boutique stores to supermarkets, was the easy part. It turns out that fulfilling even a simple request like “I need a wedding dress in Paris” is incredibly complicated. Are you a bride, a mother-in-law, or a guest? What season is it? How formal is the marriage? What statement do you want to make? Even if those questions are resolved, different AI models have different views on such things. “What we found, due to the lack of consistency and reliability of the model – and manipulation – sometimes the model will throw out one or two of the questions,” said Bornsten.. A user in the long beta test of daydream would say something like, “I’m a rectangle, but I need an outfit to make me look like an hourglass.” The model will respond by showing clothes in geometric patterns.
In the end, Bornstein understood that he had to do two things: the planning of the app for the 2024 launch of the app (although now, Daydream is still technically available in beta until 2026) and improve its technical team. In December 2024 he hired Belousova, the former CrubHub, which brought in a team of senior engineers. Daydream’s secret weapon in the fiercely competitive talent battle is the opportunity to work on an engaging challenge. “Fashion is such a fun space because it’s about taste and human creativity and visual data,” Belousova said. “It’s an interesting problem that can be solved.”
What’s more, Daydream should solve this problem twice-First by interpreting what the Customer says and by matching them with some of their quirky ways with animals on the side of the catalog. By installing the same I need to wear a repeat bat mitzvah where my ex is with his new wife, Understanding that is important. “We have an idea today to get a shopper’s vocabulary and a shopper’s vocabulary, right?” Bornstein says Bornstein. “Sellers talk about categories and qualities, and buyers say things like, ‘I’m going to this event, it’s going to be on the rooftop, and I’m going to be with my boyfriend.’ How do you combine these two vocabularies into something at startup? And sometimes it takes several Interactions in a conversation. ” Daydream learned that language is not enough. “We use virtual models, so we actually understand the products very well,” he said. A customer can share a specific color or show the necklace they will be wearing.
Bornstein says Reverdream’s next produced better results. (Even when I tried, the request for black Tuxedo pants showed me beige athletic-acture pants more than I asked for. “Each one makes a special call. We have a color, one for the fabric, one for the season, one for the location.” For example, Daydream found that for its purposes, OpenAi models are quite good at understanding the world from the point of view. Gemini’s Gemini is not yet low, but it is fast and accurate.
