Here's something most eCommerce brands haven't quite absorbed yet: their product pages are being read by two very different audiences, and only one of them is human.
A runner shopping for trail shoes sees the same static copy as someone picking out a last-minute gift. Fair enough—that's been the web for three decades. But increasingly, there's a third visitor: ChatGPT, or Gemini, or Perplexity, scanning your site for product data to serve up in some AI chat miles away from your carefully optimized checkout flow. And unlike the human shopper, the AI doesn't care about your brand story or your lifestyle photography. It just wants structured facts.
Kinect, a San Francisco startup emerging from Y Combinator's Spring 2026 batch, thinks this split is costing merchants real money. The company's answer? Build storefronts that adapt intelligently for people and publish clean data feeds for the machines. Two surfaces. One integration. Call it hedging, or call it pragmatism—either way, the founders believe brands can't afford to pick sides.
"Most companies are still thinking about their site as a single surface," says CEO Kratik Agrawal, who spent years at Google working on commerce infrastructure before stints at defense tech company Anduril and security firm Verkada. "But shopping is already splitting. Some customers browse your site. Others ask an agent to find them something. If you're not built for both, you're leaving one of those channels to chance."
What They've Actually Built
Kinect's product is less flashy than its thesis suggests. On the customer-facing side, it's an AI assistant embedded directly into product pages—fairly standard stuff at this point—but with a twist. The entire page reshapes itself depending on who's visiting and why. A first-time browser arriving from a gift guide? The system surfaces shipping and return policies upfront, along with gifting suggestions. A returning customer who's been hunting specs? Reviews and technical details move to the top. The assistant itself fields questions, handles objections, and can nudge items into the cart.
That's the surface visitors see. The other half lives off-site, and it's arguably where Kinect's real bet lies. The platform generates what the company calls an "AI Storefront"—a constellation of structured endpoints with names like brand-facts, products.json, feeds/chatgpt.json, fit.json, and llms.txt. These are designed for external agents to consume directly. When ChatGPT or some other AI assistant needs live inventory counts, fit details, or accurate pricing, it can query Kinect's own agent in real time instead of guessing from scraped HTML.
The idea is simple, maybe even obvious in hindsight: replace whatever an AI platform might hallucinate about your brand with data you actually control.
Agrawal and co-founder Varun Kandula (MongoDB alum, former consultant to Sephora and Ralph Lauren, and once a Division I squash player—a biographical detail of uncertain relevance) frame this as solving two problems at once. Product pages have been static and samey for too long. And letting AI platforms improvise your brand narrative from whatever they scrape? That's not a strategy, it's a gamble.
Landing in the Middle of a Platform War
Kinect's timing is either impeccable or merely reactive, depending on how generous you're feeling. The company's launch comes as nearly every major platform scrambles to own the agent-commerce layer.
Shopify introduced Agentic Storefronts late last year, letting merchants sell inside AI chats with native checkout and attribution baked in. Google rolled out something called the Universal Commerce Protocol at the National Retail Federation conference in January. Stripe and OpenAI co-developed the Agentic Commerce Protocol last fall. Amazon rebranded its Rufus assistant to "Alexa for Shopping" in the spring, adding features that sound vaguely dystopian—"Buy for Me," "Shop Direct"—but route users back to brand sites instead of keeping them in the Amazon walled garden.
The anecdotal evidence is piling up too. Macy's disclosed in March that shoppers using its "Ask Macy's" assistant were spending about 4.75 times more than those who didn't. Sephora launched an entire app inside ChatGPT around the same time. If you're watching this space, the signal is unmistakable: agent-mediated shopping isn't speculative anymore. It's happening, and merchants need infrastructure to show up coherently.
Kinect positions itself as the connective tissue. On your site, the assistant and adaptive pages lift conversion. Off your site, the structured feeds keep agents from inventing your stock levels or pricing.
Whether that's genuinely differentiated or just good positioning in a noisy market remains to be seen.
The Early Numbers (With Caveats)
According to the company's Y Combinator profile, Kinect had 11 live customers as of Spring 2026, spanning wellness, fashion, sporting goods, and consumer products. Early beta partners have reportedly seen conversion lifts in the 10 to 15 percent range, with shoppers who engage the assistant converting at two to three times the baseline rate. Perhaps more telling: 80 percent of assistant conversations are with first-time visitors, which suggests the tool is doing discovery work rather than just closing repeat buyers who were already half-sold.
The company's own site reports a 20 percent conversion lift, a 14 percent bump in average order value, and 24 percent more time-on-page among beta partners as of 2026—vendor-reported claims without independent confirmation. No individual case studies have been published yet, though brand logos for Slumberkins, ALC, and Huppy appear prominently on Kinect's homepage.
Most partners apparently go live within a week. Kinect runs what it describes as white-glove onboarding—joining customers' Slack channels, holding weekly performance reviews, the works. Pricing isn't public, which either means it's bespoke or the company hasn't settled on a model yet.
The team itself is tiny. Y Combinator lists two people; LinkedIn suggests somewhere between two and 10. The company is backed by YC's Spring batch, with partner David Lieb, though no other funding has been announced. Kinect's blog, active since early in the year, includes essays with titles like "The Intent Gap," "The Post-Click Problem," and "Amazon Versus the Agents"—the kind of thought leadership that signals either genuine strategic thinking or a savvy content marketing operation. Probably both.
A Crowded, Messy Field
Kinect is hardly alone in this space. DaVinci Commerce announced its "Agentic BrandStore" in March, focusing on LLM presence with support for both the Agentic Commerce Protocol and Universal Commerce Protocol. Firmly launched "Firmly Connect," a no-code platform for merchants looking to plug into agent channels without engineering lift. Runner AI markets itself as a "self-optimizing eCommerce engine." Rep AI calls its offering an "Agentic Commerce OS" with proactive engagement widgets. Wizard, Shoplazza, Commerce (which owns BigCommerce)—all have rolled out agent-related features in the past several months.
What Kinect claims as its edge, at least rhetorically, is the two-surface model. Most competitors lean heavily toward either on-site assistants or off-site agent integrations. Kinect treats both as parts of a unified system, betting that brands need to win in both arenas simultaneously, not sequentially.
It's a reasonable thesis. Whether it holds up as the definitive architecture for agentic commerce—or just one approach among many—will depend on execution, not positioning.
Betting on the Split
The founders' core belief is straightforward: shopping is bifurcating. Some people will keep browsing traditional sites, clicking through filters and reading reviews. Others will offload the work to ChatGPT, asking it to find trail shoes under $150 or pick a birthday gift for a cousin they barely know.
Agrawal and Kandula are building for a world where your storefront needs to serve both the visitor who lands via a Google search and the agent that never loads a page at all. The early numbers—limited as they are—suggest the model works, at least for brands willing to experiment.
Whether Kinect can scale beyond its initial beta cohort and into the broader Shopify-and-beyond ecosystem is the open question. But with Shopify, Google, Amazon, and OpenAI all aggressively building agent infrastructure, Kinect's timing looks less like luck and more like reading the room.
The company is taking applications for new partners. The product is live. The market, for better or worse, is moving. And for brands exhausted by static product pages and hallucinated inventory levels in AI chats, Kinect is offering—well, if not an answer, then at least a plausible direction.
