In partnership with
This week on SHIFT
Early 2026 was supposed to kick off the year you'd shop from a chat window. ChatGPT got a Shopify integration, retail behemoths including Walmart signed on, and the pitch was simple: the assistant that writes your emails would also find your shoes. Nine months later, the traffic those chatbots actually send to stores is, according to someone who can see it, not major. Among the big problems here isn't that AI can't help you shop. It's that retail runs on a pile of highly specific and manual problems, and a generalist bolted on top doesn't solve them.
Nasrin Mostafazadeh is co-founder and CEO of Verneek, which builds domain-specific AI for retailers and brands, Nordstrom among them. She's also a dear friend, and this conversation is part of our oral history project, so we start way back: a 14-year-old leading an all-girls robotics team from Iran to the world stage in Germany in 2006, where she met the boy who'd one day become her co-founder and husband. From there we get into what generic LLMs get wrong about retail and what it takes to fix the back office nobody sees.
Advertisement
You Use AI Every Day. Why don't you own any of it?
You probably opened something powered by AI before your first coffee this morning. Millions of people do. Almost none of them own a piece of it.
That's been the catch with the AI boom — the biggest gains went to insiders and venture funds, while the companies actually building the frontier stayed private and out of reach.
That's starting to change. A wave of AI IPOs is forming, and a few are shaping up to be among the most retail-accessible mega-listings in years — the rare chance to get in near the ground floor instead of reading about the gains after the fact.
Our free briefing lays out the timeline, the access window, and the filing risk most investors will miss. No credit card. Sent straight to your inbox.
Three things to listen for
1. The unglamorous job that runs retail
Every product needs its color, shape, pattern and sleeve length tagged before a merchant can buy it, stock it or sell it online. At Nordstrom that was hundreds of people doing it by hand. Verneek's system made item setup about 20 times faster and came out 20% more accurate than, in Nasrin's words, "the most caffeinated human being."
2. Nobody wants a paragraph in the peanut butter aisle
Verneek learned early, in physical stores, that even response length decides whether an AI helps a shopper. In fashion and luxury, Nasrin says, the text is the least of it: shopping is visual, and good answers look like curated outfits, not chat replies.
3. A gym teacher launched her AI career
After a year of skipping class for robotics, one teacher (phys ed, of all subjects) wouldn't pass her unless she translated a stack of English papers into Persian. She Googled for a shortcut, landed on Wikipedia's machine translation page, and followed it into natural language processing.
As a scientist, I would argue AGI itself is a myth.
Listen now
SHIFT publishes every week. If someone forwarded this to you and you'd like to get it directly, subscribe at shiftshow.ai. And if you have a question, a guest suggestion, or a story we should cover, reply to this email. Jennifer reads them all.