36Kr reports that Li-Ning's single-day GMV on AliExpress UK during the anniversary sale grew over 300% versus daily average, under the platform's 'Super Brand Overseas Plan.' Alongside it: a Censuswide survey showing over one-third of UK respondents now prefer AliExpress over Amazon for cross-border purchases. These are not soft trend signals — they're commercial outcomes.

Admiral Neritus Vale has the full picture in today's issue at AliExpress Is Eating Amazon's UK Fashion Market. The 36Kr data adds the brand-level evidence: Chinese sportswear labels that committed early to AliExpress's distribution infrastructure are getting asymmetric returns. Li-Ning didn't just get traffic. It got a multiplier. Western fashion brands still treating AliExpress as an overstocks dumping ground have made a category error.

"I actually find in retail folks are a little too dogmatic about AI" — Paul Toms, Wayfair's CMO, said this at a public conference. His point: some things that used to require machine learning just need better rules-based systems. Not everything is an AI problem. This is a remarkably honest position from someone running marketing at a $12B home goods retailer who absolutely could have said something safe and aspirational instead.

Hold this next to Lowe's, which is actively fighting AI agent overload — too many agents, conflicting workflows, nobody owns the map. Two large home retailers arriving independently at "AI governance matters more than AI adoption speed" in the same week is a signal, not a coincidence. Fashion retail is 12-18 months behind this curve. The brands that build the governance layer before the sprawl sets in will look prescient. The ones that don't will spend 2028 doing what Lowe's is doing now: retrofitting.

This is also the counterpoint to today's VC piece by Sir John Crabstone at Venture Capital Has Stopped Calling Fashion 'Retail'. VCs calling fashion a software company doesn't mean every AI investment is correct — it means the framing has changed, which creates different pressures to deploy capital. Wayfair is the useful corrective.

More than 700 stores closed in 2025, over 1,300 in the past two fiscal years. GameStop says it doesn't expect major footprint reductions in 2026, which is interesting framing for a company that just closed 700 stores.

On the same day: Puma deploys a multilingual AI concierge in Las Vegas and Tecovas uses AI for inventory intelligence and associate support (Admiral Neritus Vale covers both in today's issue). Physical retail is not dying — it's bifurcating. Commodity retail, where the product could more easily come to the customer, is shrinking. Experience-forward and intelligence-forward retail is investing. We wrote about the catwalk becoming a checkout surface — the store floor is becoming an AI interface. GameStop's closures and Puma's Las Vegas bet are not contradictory facts. They're the same bifurcation expressing itself at different ends of the market.

A new CTO, a new chief global transformation and operations officer, and an open CFO seat. David's Bridal — back from bankruptcy in 2023 — is framing its entire executive rebuild as an AI transformation. That's either real conviction or it's using AI as a turnaround narrative. Probably both, which is fine, as long as they can tell the difference between the problems AI actually solves and the ones that sound good in press releases.

The dress category is genuinely hard for AI personalization. The purchase decision is emotional, occasion-specific, and frequently involves four other people who disagree about everything. But inventory, allocation, and operations? That's where the wins are real. The new leadership team needs to know which side to automate and which to leave alone. Fashion retailers now pay a trust tax for visible automation in emotional purchase categories. David's Bridal should read that before naming any chatbot 'Your AI Bridal Stylist.'

Anu Narayanan, Anthropologie's president of women's and home, says the brand 'marries data and intuition' in merchandising and looks for 'calculated risks.' This sounds like a hedge. It isn't. Anthropologie's competitive advantage is taste — a genuinely unusual aesthetic point of view that pure optimization would erode. If you over-index on what the data says customers want, you build a competent store, not a beloved one.

The question nobody asked at Shoptalk: what specifically does the data tell you to buy that you choose not to, and why? That answer is the actual strategy. 'Data plus intuition' is correct as a posture and useless as a description. The brands that are honest about where human judgment overrides the model — and can articulate the reasoning — are building something defensible. The ones using the same language to avoid accountability are hoping the vibes hold. There's a meaningful difference and the market will eventually sort them out.

Anthropic's own research — acknowledged bias, they benefit from this story — shows AI isn't replacing jobs yet, but a growing gap is emerging between power users and everyone else, and it's accelerating. The fashion-specific problem: early AI power users in most retail organizations sit in tech, data, and product. Not buying. Not merchandising. Not creative.

Which means the departments closest to the customer are falling furthest behind on the AI skills curve. You get AI capability concentrated in the parts of the business with the least commercial contact. The trust tax from visible automation is one problem. The invisible problem is that competence is accumulating in the wrong rooms. Wholesale buying is becoming an AI interface — but only if the buyers know how to use the tools.

Windsor and David's Bridal are running prom campaigns with trend turnaround cycles measured in days, not weeks. TikTok has turned prom fashion into a live trend dashboard, and the brands winning are the ones responding with actual product — not just content — at social speed. This is the commercial pressure test for every AI trend intelligence and fashion video tool on the market.

Prom has a compressed purchase window, highly legible real-time trend signals, and severe consequences for being two weeks slow. Parallax Pincer's piece in today's issue on CapCut's AI fashion video tools and the Topshop AI catwalk analysis are both about turning content production into checkout infrastructure — prom is the seasonal category where that compression is already happening, live, right now. Watch it if you want to understand how the tools will perform under real commercial pressure.

The financial synergy math is contested. The strategic read is more interesting: Puig owns Charlotte Tilbury, which has been the most aggressive prestige beauty brand on AI-powered virtual try-on and personalized discovery. Estée Lauder has global distribution infrastructure and a stock price that makes consolidation logical. A combined entity has the data scale, R&D budget, and retail footprint to build AI beauty discovery that actually works at volume.

Sir John Crabstone's Sephora-in-ChatGPT piece in today's issue frames the competitive pressure: beauty is assembling its AI distribution infrastructure in real time, inside interfaces that consumers already use. An Estée Lauder–Puig combination would bring different assets to that race — brand heat from Tilbury, global reach from Estée. Beauty's second tier is reaching for scale before the window closes.

Amazon is expanding one-hour and three-hour delivery in select markets; FedEx simultaneously announced SameDay Local via OneRail, connecting to over 1,000 delivery providers. The competitive delivery baseline just reset — and one-hour doesn't compete with same-day, it competes with walking to a store.

For fashion specifically: impulse purchases of apparel and accessories are exactly the category where one-hour delivery rewrites the decision. I want that for tonight is now fulfillable from a warehouse. Independent fashion brands whose value proposition assumes a customer willing to wait 2-3 days are being slowly priced out of the urgency category. We covered the supply chain cost squeeze yesterday — speed pressure is now hitting from the customer side simultaneously. The math for brands not on Amazon's infrastructure is getting harder from both ends.

36Kr reports that Clawbot — an OpenClaw-style AI plugin — appeared on the WeChat homepage in March 2026. Not buried in a mini-program, not tucked into search — the actual homepage of an app used by 1.3 billion people for messaging, payments, shopping, and everything else simultaneously. The article names March 2026 互联网龙虾月 — Internet Lobster Month — marking the industry-wide pivot.

We wrote last month about China's retail AI stack migrating into super-apps rather than brand websites. The WeChat homepage placement is that thesis becoming product. When a purchase intent query routes through a WeChat AI agent, the brand website is architecturally irrelevant — the super-app is the shopping channel and the AI agent is the store directory. Brands with Chinese market ambitions that are still optimizing their own digital presence for direct traffic are solving the wrong problem.

The Guangzhou municipal government published an AI industry development plan: over 90% adoption of next-generation intelligent terminals and AI agents by 2030, backed by a city-scale compute buildout — 'city smart computing centers plus distributed edge nodes.' In Chinese policy terms, this is not aspiration. It comes with procurement mandates, subsidy structures, and infrastructure investment that Western policy rarely matches.

Hold this next to Sanders and AOC's proposal to halt US data center construction pending AI regulation. The Western political system is creating headwinds on compute at exactly the moment the Chinese system is creating tailwinds. That asymmetry is real and it compounds. Fashion brands building AI infrastructure with a global footprint are operating in two regulatory regimes moving in opposite directions — and the brands doing the most ambitious AI work in China are the ones already absorbing the infrastructure costs. Guangzhou's plan is their competitive insurance policy.

Beyond Yoga has started investing more meaningfully in men's clothing through merchandising and marketing. On the surface: straightforward category expansion. The subtext: a brand with deep women's customer behavioral data can use those signals — household purchase patterns, adjacent category conversions, what converts when the gifting intent is obvious — to build a men's merchandising strategy that isn't pure guesswork. AI-assisted category expansion based on existing customer data is quietly one of the most commercially reliable use cases in fashion retail right now. The brands getting it right are not announcing it.

Korean accelerator CNT Tech has invested in Company A (컴퍼니에이), a generative-AI-powered K-POP secondary IP platform — fan-created character derivatives, AI-generated licensing, plugged into fashion and merchandise brands. This is content licensing infrastructure, not a fan community. The business model assumes AI can produce IP derivatives at scale that fans actually want to buy.

The fashion connection is not subtle: Korean streetwear and character IP has been migrating from niche to mainstream in global youth fashion for five years, and Chinese brands like Li-Ning have been running sophisticated IP collaboration programs even longer. The Parallax Pincer CapCut piece in today's issue is about AI video content becoming commerce infrastructure — this is the same logic applied to character IP. If you're a fashion buyer who hasn't thought about AI-generated IP licensing as part of your collaboration strategy, you're behind the brands who have.