Mind the Gap: How Britain Compares to Asia on AI Commerce Adoption

A note on perspective before we start. Most of what follows leans on Thailand and Malaysia as the primary reference points, with China and India as supporting evidence. That's deliberate. I spent around five and a half years living across the region (first Johor Bahru, then Bangkok), working across both markets. The lived texture of this piece sits with the markets I can speak to honestly. The figures from China appear where they're unambiguous and publicly verified, and they're useful precisely because they're unambiguous and publicly verified.

---

The clearest evidence I've ever seen of how different the conversational-commerce infrastructure is between Britain and Southeast Asia did not come from a report. It came from a WhatsApp group.

A colleague in Kuala Lumpur had me added to the chat for her building's corner minimart. Twenty-six neighbours, the shopkeeper, and a running thread of orders that read like a weather forecast. You'd post your list before lunch. The owner would emoji-react and screenshot a printed total. A bag with your name on it would appear outside your flat door later that afternoon. There was no app. There was no checkout. There was a chat, a reasonable mutual trust, and a PayNow QR code at the bottom of the shopkeeper's daily summary.

When AI shopping assistants arrived in that environment, they didn't have to convince anyone that commerce-in-conversation was a legitimate idea. The idea was already where the customers lived. It just had to get smarter.

Britain is trying to bolt the same idea on to an infrastructure that wasn't built for it. That's the gap this piece is about, and it's a more durable gap than the adoption-rate league tables suggest.

The adoption gap is real, with some asterisks

Let's start with the headline figure, because it's worth getting right. ONS Wave 141, the late-September 2025 release, puts UK retail AI adoption at 17%. That's against 23% across all UK businesses, and 36% for large companies. The retail-sector figure has been climbing slowly; the September read is up from earlier in the year. But it's still the lowest sectoral cut in the bulletin.

The cross-market comparison is where this gets messy, and it's worth being explicit about the methodology because the original version of this piece wasn't as careful as it should have been about which numbers measure what. The cleanest cross-country benchmark for active enterprise AI use is the IBM Global AI Adoption Index 2023, which surveyed IT professionals at large organisations across fifteen markets. By that measure, organisations in India (59%), the UAE (58%), Singapore (53%) and China (50%) lead. The UK didn't make the top tier. On the related question of who is most likely to accelerate AI rollout, IBM puts China at 85%, India at 74%, the UAE at 72%, and the UK at 40%, among the lowest in their sample alongside Australia and Canada.

The ONS and IBM figures aren't strictly comparable. ONS asks broadly across the whole business population; IBM samples IT professionals at larger firms. The methodologies differ. But the direction of travel is the same one Microsoft and Stanford have been reporting independently, and the direction of travel for the UK is not flattering: behind the markets we'd expect to be ahead of, and behind some we wouldn't.

Direction of travel matters more than the exact decimal points. The decimal points are debatable. The direction isn't.

What's underneath the numbers

Adoption rates measure a state, not a story. The story is what makes the state plausible, or implausible, to begin with.

In Thailand, LINE has 54 million users, accounting for 77% of the country's mobile internet users. The average user spends 67 minutes a day on the platform, out of 216 minutes total on their phone. Sixty-seven minutes is more than many people in the UK spend on every messaging app on their phone combined. But LINE in Thailand isn't really comparable to WhatsApp here. It's somewhere between Facebook, WhatsApp, your debit card, and the QR-code menu at every restaurant. All at once.

The business layer of this is the thing British retailers tend to underestimate. There are more than three million LINE Official Accounts: businesses, government services, banks, brands. 92% of Thai internet users open LINE at least weekly. Over 85% of Thai shoppers interact with a brand through its LINE OA before visiting that brand's website. The OA isn't the marketing channel; it's the storefront. The website is where you go to look up a product spec when the OA conversation has already convinced you.

The picture in China is the same idea at an order of magnitude greater. WeChat Mini Programs generated approximately ¥3 trillion in GMV in 2024, around $400 billion. For scale, the whole of Southeast Asia's platform e-commerce GMV came to $157.6 billion in 2025 (Momentum Works). One product, inside one Chinese super-app, runs at over two and a half times the combined platform commerce of the entirety of SEA. And the format of that ¥3 trillion is, again, conversational and embedded: you don't leave the chat to make the purchase; the mini-program runs inside WeChat, with payment via WeChat Pay one tap away.

Underneath all of this sits a class of real-time payment rail that the UK has, in different form, had since 2008, but doesn't use in the same way. PromptPay, Thailand's instant interbank transfer system, processes around 74 million transactions a day. Total registered users (across multiple bank accounts and phone numbers per person) sit above 90 million in a country of about 71 million people. Singapore's PayNow has equivalent reach for its population, as does Malaysia's DuitNow and India's UPI. The interesting thing isn't that these rails exist. It's that consumers use them by default: for splitting dinner, for paying the corner shop, for closing out a LINE OA order. They're not "a payment option"; they're the substrate.

Faster Payments in the UK is technically excellent and underused for consumer transactions. Most British shoppers, when ordering from a chat with a small retailer, would still reach for a card.

The British comparator is fragmented

The UK has a conversational-commerce infrastructure of sorts. WhatsApp Business is genuinely useful for small retailers, and adoption is climbing. Meta reports over 400 million monthly active users globally, of which a majority are SMEs. But the closest thing the UK has to LINE OA's near-universal commerce coverage is fragmented across WhatsApp Business at small scale, Instagram DMs at a different scale, and email at older-school scale. None of them has the chat-as-storefront defaultness that LINE has in Thailand or WeChat has in China.

That's not a brand failure. It's a coordination problem with regulatory, infrastructural, and habitual layers. The UK never developed the super-app pattern, partly because the platform commercial environment didn't push that way, partly because the regulatory environment is structured around platform fragmentation rather than consolidation, and partly because consumer habit was already settled around browser-and-app patterns by the time messaging platforms became plausible commerce surfaces.

This means that when an AI shopping assistant arrives in a UK retail context, it has to do two jobs at once. It has to be smart enough to be useful, and it has to convince the customer that "having a conversation with a retailer" is a normal thing to do. In Thailand, only the first of those jobs needs doing.

The May 2026 punctuation

Two announcements from the same week of May 2026 illustrate the point, and they were initially the headline I wanted to hang this piece on. They turned out to be more useful as punctuation than as spine.

On the 11th, Alibaba announced that its Qwen AI app would integrate with Taobao and Tmall, giving the agent direct access to their four-billion-item catalogue: text or voice search, comparison, virtual try-on, 30-day price tracking, and checkout via Alipay, all inside a chat thread. Two days later, Amazon retired Rufus and launched Alexa for Shopping in the US, with a "Buy for Me" feature that completes purchases on third-party retail sites. Two of the largest commerce platforms in the world shipped broadly equivalent capabilities within forty-eight hours of each other. Different press receptions, different framings, but the underlying technology bet was the same.

The interesting bit isn't who announced first. Both Alibaba and Amazon got there at roughly the same time, and given how integration timelines work, both have been building this for at least two years. The interesting bit is which user base each platform is launching into. Alibaba's Qwen is rolling into a Chinese consumer base that has been buying via in-chat super-app interfaces for the better part of a decade. Amazon's Alexa for Shopping is rolling into a US consumer base that has been buying via search-then-page-then-cart patterns for nearly thirty years. The same product faces a very different first ninety days depending on which side of the conversation-vs-browser divide the customer arrives from.

There's a related observation worth surfacing. A May 2026 ResearchAndMarkets report puts global consumer AI use at 62% for product comparison, 23% at checkout, and 19% post-purchase. The gap between discovery and execution is not a marketing problem. It's an authorisation, infrastructure, and trust problem. Engineering work, not messaging.

62%

23%

19%

Comparison

Checkout

Post-purchase

The economies that close that gap fastest are the economies whose infrastructure was already pointed in that direction. Asia's chat-and-rails pattern reduces the friction in checkout because the rails were already trusted and the chat was already where commerce lived. The UK's gap will close more slowly, because closing it means rebuilding consumer habit at the same time as deploying the technology.

What this means for British retailers

A few things, none of them ground-breaking, all of them worth being honest about.

1. The platform layer is becoming the discovery layer here too. Google's Universal Commerce Protocol has a Buy button live in US Search via Wayfair. Amazon's Alexa for Shopping is the equivalent move from a closed ecosystem outward. The pattern arriving in Britain over the next eighteen months looks structurally similar to the SEA platform-native model, just with different platforms running it. If your direct-to-customer relationship depends on owning the storefront, the storefront is moving.

2. Discovery without checkout-readiness is a leaky funnel. The 62%-to-23% drop-off measured by ResearchAndMarkets is what you should be reading as your engineering backlog. If an AI-mediated shopper finds your product, can they complete the transaction without their AI failing them? The answer for most UK retailers, today, is "depends on which agent and which checkout flow", which is a polite way of saying probably not. Building this isn't a marketing project.

3. The conversational-commerce habit is going to arrive eventually, but it won't arrive on platforms anyone here owns. WhatsApp Business is the closest infrastructure we have. It will probably do most of the heavy lifting in the SMB segment for the next few years. Larger retailers will face a tougher choice: build for messaging-as-storefront with the platforms that exist, or wait for a UK-native pattern that may not come.

A note on what I'm not saying

I'm not saying British retail technology teams aren't sophisticated. Several of them are excellent. I'm not saying the SEA model transplants cleanly to the UK; different payment habits, different regulatory environment, different consumer expectations around data and privacy all complicate the transfer. And I'm not predicting the gap closes on any particular timetable.

What I am saying is that the gap is real, the gap is structural, and the gap is not principally about technology. It's about consumer habit and the infrastructure beneath consumer habit. AI commerce arrived into a different set of starting conditions in Bangkok and Shanghai than it did in Manchester or Reading. Pretending otherwise is comforting but doesn't help anyone build for the next eighteen months.

Worth paying attention to. Worth being honest about where we are. Worth being clear that "we'll get there" isn't a strategy until we know what infrastructure we're getting there on.

---

Data sources and further reading

• ONS Business Insights and Conditions Survey Wave 141: AI adoption in UK business (October 2025)
• IBM Global AI Adoption Index 2023: enterprise AI use by country
• LY Corporation: LINE Thailand (December 2025)
• LY Corporation: LINE for Business in Thailand (January 2026)
• Tech Buzz China: WeChat Mini Programs e-commerce GMV (2024)
• Nation Thailand: PromptPay 2024–2025 volume per Bank of Thailand
• DealStreet Asia: Momentum Works Ecommerce in Southeast Asia 2026 report
• Campaign Asia / Mintel: Thai consumer AI shopping attitudes (2025)
• Yahoo Finance / Reuters: Alibaba to integrate Qwen AI with Taobao (May 2026)
• Amazon: Meet Alexa for Shopping (May 2026)
• ResearchAndMarkets: AI Shopping Agents and Agentic Commerce 2026 (Yahoo Finance reprint)