AI Commerce Weekly: Week 26, 2026
Three UK studies say shoppers won't hand AI the wallet. The supply side spent the week laying rails anyway, and fencing Britain out.
TL;DR
This was the most British week the feed has run in a while, and it split cleanly in two. Three UK research houses, dentsu, PSE and Quickfire, all reached the same verdict inside seven days: shoppers will happily let AI help them discover and compare, but most will not hand it the wallet. The supply side ignored the caution and laid rails anyway. Sainsbury's built its retail-media engine in-house while Asda rented Amazon's. Disney, Walmart and Newegg all shipped assistants, with Newegg's grounded so it cannot invent stock, and walled off from British shoppers on day one. The gap between what buyers permit and what retailers build is the whole story.
The demand side won't hand over the wallet
Three separate British research houses published inside the same week, and they landed on the same finding from three different angles. dentsu UK&I surveyed 2,003 UK consumers and found more than a third now use AI shopping tools, rising to over half among Millennials and Gen Z, with almost 40% using AI to find the best product, compare brands and hunt cheaper alternatives. Then the brake: nearly two-thirds remain uncomfortable letting AI purchase autonomously on their behalf. The signals that anchor the final decision are stubbornly old-fashioned, customer reviews (54%), friends and family (48%) and in-store information (45%). dentsu's own framing is the one to keep: now that discovery is everywhere, the scarce commodity consumers actually want is guidance.
PSE Consulting reached 4,250 AI-shopping consumers across the UK, US, France and Germany and found 74% prefer an independent AI assistant for discovery, a universal tool like ChatGPT or Gemini (41%) or a specialist category assistant (33%), while just 10% want AI baked into a retailer's own platform. Ninety per cent expect their marketplace use to hold or grow as AI adoption rises. PSE's Chris Jones reads it as consumers treating discovery and execution as distinct stages, leaning on established brands for fulfilment, payments and operational trust while the competitive fight moves upstream into the discovery layer. That is a more useful mental model than the "agents will disintermediate everyone" line, and worth sitting with before you assume the assistant is a threat to your marketplace rather than a new shop window onto it.
The third study is the one that should worry a technology leader most, because it points the finger inward. Quickfire Digital surveyed 2,028 UK consumers and 201 retailers and found 45% of shoppers frustrated with AI-powered experiences, citing generic recommendations and, tellingly, AI suggesting out-of-stock items. On the retailer side, 84% said planned improvements were delayed or never delivered in the past year, and 40% now spend more than a quarter of their e-commerce budget just maintaining existing systems. Co-founder Martin Harper calls it a hidden "Growth Tax". The same study found the owned website still beats social nearly three to one for discovery (45% versus 16%). Sarah has written before about the cost of mistaking consumer enthusiasm for consumer trust, and this is the week's cleanest evidence of the gap: the appetite for AI help is real, the willingness to delegate is not, and half the retail estate cannot ship the improvements that would close it.
Retail media, build it or rent it
The cleanest strategic split of the week was two UK grocers answering the same question in opposite ways. Sainsbury's retail-media arm Nectar360 reported early results from Pollen, a unified retail-media platform it built in-house to attack the sector's fragmentation problem. Pollen pulls audience insight, planning, activation, in-flight optimisation and measurement into a single environment with multi-touch attribution at its core, and uses AI throughout, including a creative-optimisation tool that scores assets against years of historical data (in test with PHD UK and Lipton). The early figures are eye-catching, over 2.5x higher incremental sales from omnichannel campaigns, up to 10x higher conversion from first-party-data targeting, and up to 25% of incremental sales from mid and upper-funnel activity, with Unilever and Coca-Cola Europacific Partners named onboard. Those numbers are first-party and vendor-reported, so treat the multiples as directional rather than audited. The point that matters is architectural: a UK grocer chose to build the measurement-and-activation stack itself rather than rent one.
Two days earlier, Asda went the other way and signed a retail-media partnership with Amazon Ads, deploying Amazon Retail Ad Service across both Asda.com and its George fashion and home label. Amazon's Joseph Park called it "the first of its kind in the UK". ARAS uses machine learning and shopping-behaviour data to serve intent-based sponsored ads, and lets brands manage campaigns across Asda's inventory and Amazon's through one framework, with a phased rollout from Q4 2026. Asda's Rachel Eyre framed the ambition as making Asda "the most effective and frictionless place for brands to reach British shoppers". So there it is, side by side: Sainsbury's owning the engine and the data ontology, Asda renting Amazon's proven one and accepting the dependency that comes with it. Neither answer is wrong. The George angle makes the Asda deal a direct fashion-retail data point, and the build-versus-rent decision is the same one every retail-media team will face this year, just rarely this legibly.
Grounded beats glib, and Britain gets fenced out
Three conversational assistants shipped this week, and the instructive one came from the US. Newegg launched a full conversational shopping assistant at newegg.com/ai that uses retrieval-augmented generation, querying its live catalogue at every turn for real-time price, stock and deals before it says a word. That is a direct swing at the failure mode Quickfire's shoppers were complaining about in the section above: assistants confidently recommending things that are out of stock, mispriced or incompatible. For a multi-constraint buy like PC parts, grounding in live data is the difference between a toy and a tool. Newegg also shipped a ChatGPT app built on OpenAI's Model Context Protocol, so a shopper can pull live Newegg inventory inside ChatGPT with checkout bounced back to Newegg.com to keep payment data out of the chat. The catch earns this a UK mention: that ChatGPT app excludes the EEA, Switzerland and the United Kingdom. British shoppers get the on-site assistant, not the ChatGPT integration. It is becoming a pattern worth naming, US-first agentic launches routinely carve out the UK and EU on day one, and if your strategy assumes British consumers are seeing the same surfaces as American ones, they often are not.
The other two assistants were brand-led. Disney Store unveiled an AI Personal Shopping Assistant, piloting in its iOS app, that mirrors Disney's brand voice and character knowledge and recommends by character, age, occasion or budget with one-click shopping and order tracking, part of CEO Josh D'Amaro's AI push. It landed alongside news of new Disney Store concessions in Selfridges Birmingham and Selfridges Manchester Trafford from mid-July, which gives a first-party, character-led assistant direct UK relevance. And in the US, Walmart embedded its Sparky assistant into a Walmart Live shoppable event for the first time during its Summer Deals campaign, a marker of conversational AI moving into live commerce that any UK retailer experimenting with live shopping will want to watch. Three different bets on the same surface: Newegg on grounding, Disney on brand voice, Walmart on live entertainment. The one to copy first is Newegg's, because grounding is the requirement the other two quietly depend on.
Being bought, not just recommended
If the demand-side research says shoppers use AI to discover, the obvious follow-on question for a brand is whether the AI can actually find you. Nudge launched an Agentic Commerce Platform alongside a $1.1m pre-seed built to answer exactly that, backed by s16vc and Antler plus operators from Shopify, Nutanix and Postman. The platform lets consumer brands measure their AI visibility (how shoppers describe them in AI conversations and which products surface in answers), enrich catalogues so agents can recommend them, and close the gap between being discovered and being bought. Co-founder Kanishka Thakur puts it plainly, and it is a good line to pin above any GEO roadmap.
Early customers span apparel, beauty, food and beverage and wellness, and the funding is earmarked partly to deepen integration with agent protocols ACP and UCP. It is a small, early round, so treat it as a signal rather than a result. But the signal is the one we have been tracking for months, most recently through Koddi's commerce-media data the week before: machine-readability is quietly becoming a sales channel. Structured product data, transparent pricing and live inventory are what let an agent recommend you, and they are the same feed quality the on-site assistants in the previous section need to work at all. One body of work, several payoffs.
Bodycare bets its comeback on the data layer
The boldest UK bet of the week came from a chain that had gone under. Bodycare has picked retail-tech firm PMC to build the digital backbone for its return to the high street, after being bought out of administration in late 2025. This is not a refurbishment, it is a ground-up rebuild: 25 stores in 2026, first openings in Leeds, Bradford, Sheffield and Derby, scaling to 200 over five years. The part worth a technology leader's attention is the architecture. Bodycare wants to run daily store operations through a proprietary AI platform, with AI decisioning shaping colleague actions off data harvested from task management, in-store networks, CCTV, point of sale and digital-shelf systems. PMC supplies its Graphene Engage platform for POS and its Connect product as the orchestration layer feeding the central AI engine, and each store gets a content-production studio for hyper-localised social made with staff and customers.
Strip away the "first AI retailer" marketing (that is trade framing, not an audited claim) and what is left is genuinely worth watching. A Northern-rooted chain, the first-wave towns are all northern, is treating the data-and-decisioning layer as the product rather than the bolt-on. It is the inverse of most legacy transformations, which graft AI onto an existing estate and its existing mess. Bodycare gets to design the data ontology before the first till is plugged in, which is the luxury most of us never get. Whether "AI decisioning powering colleague actions" survives contact with a Tuesday-morning rush in a Bradford store is the open question, and it sounds clever right up until a peak-trade incident. But the premise, build the orchestration layer first and let the AI reason over clean store telemetry, is the right shape. One to revisit when the doors actually open.
Culture is the other half
The quietest item of the week is the one that generalises best. British clothing brand Boden held its first ever internal AI Roadshow, "Behind the Screens", a 90-minute event where teams set up stalls around the Great Hall and demonstrated how they actually use AI day to day, across design, supply-chain planning and data tools. The emphasis was on real working examples rather than future potential, staff moved between stalls swapping ideas, founder Johnnie Boden toured every one, and Makers CEO Claudia Harris joined a fireside chat on AI adoption and the future of work. It is self-reported and light on hard metrics, so it will not trouble the news wires. But it is a useful counterweight to the week's vendor announcements, because it treats AI adoption as an internal-capability and culture problem rather than a procurement one.
That is the half most transformation plans underfund. It echoes the line Google Cloud gave Britain the week before, that technology is only half the answer and people are the other half. Bodycare is betting on the data layer and Boden is betting on the people layer, and a serious programme needs both. A mid-market fashion brand making AI use legible and shareable across its own teams is a template more retailers could copy for the price of an afternoon and a Great Hall.
Sources
The demand side won't hand over the wallet
All three are agency- or consultancy-commissioned surveys relayed by RTIH; attribute figures to the research house.
- New dentsu research reveals how Brits are using AI, social media and stores in shopping journeys, Retail Technology Innovation Hub, 22 June 2026
- Shoppers sign up for AI product discovery, but still look to marketplaces to complete their purchase (PSE Consulting, 4,250 consumers), RTIH, 24 June 2026
- New Quickfire Digital research reveals how AI and legacy technology are failing e-commerce (2,028 consumers / 201 retailers), RTIH, 22 June 2026
Retail media, build it or rent it
Pollen's 2.5x / 10x / 25% figures are first-party and vendor-reported; treat as directional. Amazon's Joseph Park quote is 'the first of its kind in the UK'.
Grounded beats glib, and Britain gets fenced out
- AI Shopping Assistant Launches at Newegg: Real-Time Catalog Powers PC Build Advice, Tech Times, 27 June 2026
- Disney Store gets personal as company takes wraps off new AI Personal Shopping Assistant, RTIH, 24 June 2026
- Walmart's Sparky AI assistant makes live shopping debut as retailer launches Summer Deals campaign, RTIH, 22 June 2026
Being bought, not just recommended
Bodycare bets its comeback on the data layer
The 'first AI retailer' line is trade framing, not an audited claim. This source does not name the buyout's backers.
Culture is the other half
Stay Connected
Follow LLCommerce on LinkedIn
Get the latest AI commerce insights, analysis, and industry news delivered to your feed.
Large Language CommerceAuthors
