Topshop, currently existing as a digital brand inside ASOS, staged an AI-generated catwalk and wired a TikTok live shopping stream onto it — Shark Beauty and LOOKFANTASTIC providing the purchase layer. The aspiration and the checkout button now occupy the same moment.

The architecture matters more than the execution: AI generates fashion content at scale, live commerce captures impulse in real time, and a brand without physical space gets cultural visibility at near-zero production cost. ByteDance priced AI content production at $0.14 per second in March. Someone has now wired that cost structure to a till. China's short-drama market arrived at this format eighteen months ago. London is catching up.

ASOS has the conversion data to know whether this worked. If it did, every brand without physical presence has a new playbook — and every brand still paying for a physical runway has a question to answer about what, exactly, they are paying for.

Same week, two very different bets: Pinterest launches live shopping; Walmart acquires a conversational commerce platform. Pinterest's move is the more coherent one — its AI shopper already starts from aesthetic intent rather than query, and live video feeds naturally into that visual-first discovery architecture. Walmart is buying conversation because it cannot out-Amazon Amazon on logistics, and it knows it. Whether a conversational layer actually changes the margin math for a big-box retailer is still an open empirical question.

Virgil Abloh's estate opened Off-White's first India flagship in Bengaluru — the tech capital, not the financial capital. The luxury industry reflex is always Mumbai or Delhi. Bengaluru is younger, more globally mobile, and more likely to contain the customer who already owns Off-White from a London trip and wants to buy the next drop at home. This is a customer-targeting decision dressed as a location decision. It is the right call.

Michael Kors launched an AI-powered retail assistant on its website — in the same week FashionUnited mapped Capri Holdings' ongoing post-Versace stabilisation. The two stories belong together. Capri sold its most powerful brand, kept Michael Kors and Jimmy Choo, and is navigating what a mid-market American luxury group looks like without Versace's gravitational pull.

An AI assistant is a reasonable answer to "how do we convert site traffic without adding headcount." It is not a brand strategy. Only two of seven chatbot quality dimensions actually predict a sale, and neither of them is "having a chatbot at all." Capri's real problem is differentiating Michael Kors in a squeezed mid-market — an AI assistant helps at the margin, not at the core. Today's piece on wedding planning chatbots shows where conversational commerce genuinely moves the needle; the Michael Kors deployment does not obviously share that context.

AI e-commerce startup Onton raised $7.5 million. The timing is unfortunate. Shopify just absorbed AI-powered discovery into its platform layer, available to millions of merchants at no incremental cost. When the infrastructure player commoditises your value proposition, the runway clock starts ticking — loudly.

The venture math requires Onton to either solve something Shopify explicitly cannot solve at platform scale, serve a segment Shopify cannot reach, or get acquired before the gap closes. Yupp.ai burned through $33 million and shut down nine months post-launch for approximately the same reason — the market it built for moved faster than its runway. $7.5 million is not a lot of time. The OECD's twin-transition framing for small retailers, in today's piece, maps the same squeeze from the demand side: the retailers Onton wants to serve are already being asked to absorb too many simultaneous shifts.

German trade publication etailment.de is running a detailed guide on how retailers should prepare for the AI-agent shift — framed as operational present, not future scenario. The key sentence: Nicht Menschen suchen und klicken, sondern eine Software recherchiert, vergleicht und verhandelt — "not humans searching and clicking, but software researching, comparing, and negotiating."

That word verhandelt is the gap nobody has designed for. When the agent can negotiate price or terms on behalf of the consumer, every promotional markdown strategy in retail has to be rebuilt from the ground up. We catalogued 14 documented multi-agent failure modes in April — negotiation as a default consumer behavior was not among the scenarios retailers in that study had designed for. The German piece names the gap without quite measuring its depth. It is still ahead of most of what the English-language trade press has published on this.

German e-commerce giant Otto grew GMV 6 percent last fiscal year to roughly €7.5 billion — a solid result in a flat European retail market. The number I keep coming back to is Otto's MOVEX platform generating AI model imagery at 60 percent lower cost with no disclosure standard attached to it. GMV growth and AI-generated model deployment are running in parallel; nobody has yet asked whether one is driving the other, or what consumers think they are buying when they do not know the image was generated.

Third-party Amazon sellers represent 5.5 percent of all UK retail sales. Not Amazon's own revenue — the independent sellers running on its marketplace. When that proportion of a national retail economy flows through a single logistics and discovery layer, Amazon's FBA surcharge earlier this year stops being a logistics story and becomes an infrastructure pricing story. The sellers had nowhere to go then. Shopee is running the same playbook simultaneously in Southeast Asia — raising fees in a market it controls, counting on the absence of a credible alternative. Platform dependency is not a mistake retailers made. It is a structure that was built around them.

RMB 1.5 trillion in China e-commerce. That is the backdrop for everything happening there right now: Alibaba's RMB 10 billion profit compression in service of quick-commerce dominance, Douyin's one-sentence shopping deployment, and today's Pinduoduo AI drama bet. None of these are experiments. They are capital allocation decisions in a market too large for any player to concede.

Connection: Pinduoduo's AI drama investment and the Bain projection belong in the same sentence. China is treating content as commerce infrastructure — the drama is not marketing for a product, it is the distribution channel itself. The RMB 1.5 trillion number is partially a content investment return. Bain did not frame it that way. They should have.

Reale Actives launched, sold out, and collected polarizing reviews within hours — and the CEO told Glossy they had prepared for a mixed response. That is a tell. When your launch strategy is built around surviving backlash rather than earning product trust, you have already answered the question of what the brand is for.

The creator-to-product conversion problem is consistent: audiences follow personality, not formulation credentials. Connection: This is the same fault line running through today's E.l.f. piece. E.l.f. has explicitly built its AI strategy around attention rather than conversion, on the thesis that product trust follows when the attention is right. Earle has the attention. Whether Reale Actives earns the trust of the retinol buyer — as distinct from the fan buyer — is a different test. Those are different customers, and they have different tolerances for disappointment.

Glossy's deep look at the dupe fragrance category asks the right question at the right moment. We covered Dossier's ascent to Walmart's top-selling fragrance last week — a $49 bottle with the same accords as a $300 original. The arbitrage worked so well it collapsed its own premise: when the dupe is at Walmart, the consumer advantage of discovering a secret alternative disappears, and the brand has to compete on something other than price-versus-prestige.

The options narrow fast. Build brand equity — expensive and slow — or use AI-assisted development to stay ahead of the original brands' discovery cycles, which now have access to the same tools. TikTok Shop already showed that fragrance sells through storytelling and community trust, not smell. The dupe brands that own that distribution infrastructure own the category next cycle. The ones waiting for another arbitrage window may be waiting a while.

Glossy traces the beauty bag charm trend — small branded accessories becoming cult collector objects carried in public. Parallax Pincer goes deep on the mechanics today in "The Charm Is the Product Now." The short version: the charm is a walking advertisement the consumer chose to carry, which is categorically different from every form of paid placement. You cannot buy that. You have to earn it. In a moment when every brand is running personalization engines and retail media buys, the brands winning charm loyalty are doing it through cultural intuition — not algorithmic optimization.

Flipkart elevated Sharon Pais to lead Myntra at an interesting moment: Myntra's M-Now service captured 10 percent of orders in active markets within a year, stocking 100,000 styles for sub-30-minute delivery. That is a structurally unusual bet — quick commerce works for groceries because the category is predictable; fashion SKU proliferation makes 30-minute delivery genuinely expensive to run at scale. Pais takes the wheel at the moment the thesis most needs testing.