When Meta launched Muse, its consumer AI agent, in partnership with payments processor Stripe in September 2026, it was “the watershed event” for agentic shopping, said John Harmon, senior retail technology analyst at Coresight Research. AI companies are becoming increasingly interested in helping consumers shop, and with good reason: if users can buy goods through their platforms, they can take a small cut of each transaction for a nice profit, as Meta said it plans to do. But agentic shopping isn’t simply theirs for the taking.
Searching for products and even checking out through agents like Muse and OpenAI’s Dots is just one piece of the agentic shopping puzzle. E-commerce giants like Amazon and eBay are offering their own shopping agents, and retail brands are starting to embed agentic capabilities into their own sites (and in some cases, banning agents like Muse). There are also numerous startups building in the space, many of them founded by e-commerce veterans who know the landscape well. Some like ReFiBuy and PayOS are building the infrastructure, while others like Daydream, Onton and Wizard are competing with their own agentic shopping platforms for consumers.
Deloitte projects 25% of global e-commerce sales will be enabled by AI agents by 2030, calling agentic commerce “a structural change in how demand is created, managed, and fulfilled.”In a market this crowded, where can startups truly compete?
Building the bridge to the customer
From AI platforms to marketplaces and retailers, companies competing in the space are all vying to get in front of consumers. For ReFiBuy, co-founded in 2025 by e-commerce veteran Scot Wingo, it’s an opportunity to carve a lane it can own.
The startup helps brands and retailers in consumer packaged goods, beauty and fashion optimize their product catalogs for discovery by AI chatbots and agents, efficiently populating product cards with accurate product attributes and pricing. It’s similar to what his previous company ChannelAdvisor, which he took public in 2013, did to help brands reach customers on marketplaces like Amazon. Wingo said he’s now competing against his former company as it also takes on the problem for the agentic era.
Scott Wingo, ReFiBuy CEO. Credit: Scott Wingo
While some firms are offering AEO/GEO (essentially SEO, but for being discovered by AI rather than traditional search) to help brand names get cited by AI tools and drive positive sentiment, ReFiBuy sits further down the funnel, making the actual purchase possible.
“We're trying to not only get a mention, but rather to ring the register,” said Wingo.
Julie Bornstein, another industry veteran who built up the e-commerce businesses at Nordstrom, Sephora and Urban Outfitters and was COO at Stitch Fix, is making a different play for the customer with her startup Daydream, which she launched in June 2025 after raising a massive $50 million seed round the year prior.
Daydream is an agentic fashion shopping platform that lets consumers search 10,000 brands through one chat — like a ChatGPT specifically for discovering and buying clothes and accessories. With Daydream, she’s competing directly against all the chatbots and agentic search tools as well as major marketplaces, and she sees breaking through the noise as the biggest challenge. Her strategy is to put Daydream everywhere customers may be, creating as many ways to reach them as possible.
Credit: Daydream
For example, the startup launched a plug-in for ChatGPT and is using Apple Intelligence to enable consumers to find shoppable matches for outfits saved in photos in their camera rolls. It’s also making deals with retailers to embed its agentic search technology into their websites, turning what could be a competing way for users to shop with AI into yet another customer touch point for Daydream.
“Our thought is we want to be wherever you are,” said Bornstein.
Owning the niche
By focusing solely on fashion, Bornstein believes Daydream can capture its own piece of the ecosystem. Harmon echoed this sentiment, stating that startups can leverage tools that have already been built and operate with lean teams, which can work well for serving a niche.
Fashion is a deeply personal space where everything from your body type to your age, where you live and what you do for work and in your personal life dictate what’s relevant to you, said Bornstein, emphasizing why it's crucial to understand the user. It’s also deeply nuanced with a practically infinite number of objective product characteristics (think of all the colors, sizes, sleeve lengths, jean fits and beyond) and subjective characteristics (like if an item works for a specific type of event, vibe or even body type).
As shoppers see new trends being named on social media and become more accustomed to conversational interactions with AI and search tools, consumers are now constantly gaining new language that they didn’t have previously.
Julie Bornstein, founder and CEO of Daydream. Credit: Daydream
“There are all sorts of new inputs that are constant and changing quickly. The input on things that inspire you has never moved faster,” said Bornstein, adding that Daydream is gathering enormous volumes of prompts in the fashion space and constantly working to understand the “what’s next,” while general AI tools focus on broad use cases.
ReFiBuy is similarly focusing on a niche task that companies in the space need but can’t easily execute themselves.
In the marketplace days, Wingo said, they worked with five attributes (size, color, brand, availability and price), and populating them correctly at scale was tedious enough. Now for agentic search, it’s more in the realm of a hundred attributes, as consumers use an average of 24-word prompts and four conversation turns to really refine what they’re looking for, according to Wingo. ReFiBuy aims to be useful both to the retailers who want to get in front of chatbot and consumer agent users, and to the AI platforms that want to facilitate those purchases.
Harmon also believes AI platforms won’t likely be the end-all-be-all for agentic shopping, giving startups and retailers a real shot. There are consumers using loyalty programs as well as challenges like returns and agents buying the wrong thing.
“The chatbots want to do a whole lot of other things too, like health information and business and mathematics research,” he said. “I don't think they can be everything for everybody, and that's what the retailer is really good at.”
