Why AI Is Reshaping Commerce Faster Than You Think: Highlights from Ecommerce Expo

It was a thrill to present to a packed room at this year’s Ecommerce Expo, which wrapped up yesterday. Judging by the amount of furious note-taking I saw in the audience, most people probably have enough notes already, but I wanted to share a few of the key points here too, especially for anyone who couldn’t make it.

From Crumpets to Conversions: AI’s Impact Is Already Here

I opened the session with a small story. Last week, I asked ChatGPT which toaster I should buy. I expected a quick price comparison, but instead it warned me that the one my wife liked, purely because it matched the kitchen cabinets, had a habit of welding itself to crumpets. Not the feature we were looking for, but it’s saved a few breakfasts.

It’s a silly example, but it shows just how deeply AI is already woven into everyday life. It’s in our homes, our decisions, and our customer journeys. The real question now isn’t what AI can do, it’s how you make it work for your business in a way that actually makes a difference.

And we’re already seeing results. Adobe Analytics tracked a 1,200% surge in traffic from ChatGPT searches to retail sites in the first half of this year. Those visitors stay longer and are twice as likely to convert. It’s because they’re not arriving casually anymore. They’ve done their research and they’re ready to buy.

Zalando is a great example of how to meet that expectation. After launching its GPT-powered fashion assistant late last year, it saw a 23% rise in product clicks, a 40% increase in wishlist adds and a 7% drop in returns. People aren’t just browsing with more confidence; they’re keeping what they buy.

Quiet Efficiency and Bold Moves Behind the Scenes

Growth is only part of the story. AI is also becoming a powerful way to protect margins. Dunelm, for example, has cut its “zero results” searches to under 1% by using AI to guide shoppers to the right products, reducing friction and support costs along the way. Sweaty Betty is taking a similar path behind the scenes, using a virtual analyst to piece together signals from reviews, service queries and other feedback. Spot a sudden spike in talk about early-morning runs? It won’t just flag it – it’ll build audiences around that behaviour so campaigns can land where they’ll make the most impact.

AI is also shaping the back end of retail. Wayfair uses it to tag millions of products for style and compatibility so customers can easily find matching items. Boohoo and PrettyLittleThing use it to set prices based on demand, a powerful but sometimes unpopular tactic when prices shift between clicks. Even Morrisons uses a Gemini-powered product finder that guides shoppers straight to the shelf while generating behavioural data to improve merchandising. AI isn’t just a checkout assistant anymore. It’s already helping with everything from merchandising to planning to pricing.

When Personalisation Actually Feels Human

One of the most exciting shifts is how AI makes personalisation feel more human. Stitch Fix, already ahead of the curve a decade ago, recently launched a new AI style assistant built on 15 years of data. Customers can chat for instant recommendations and still message a human stylist directly to ask, “Can I really pull this off?” It doesn’t feel like an algorithm choosing clothes. It feels like a conversation.

Not every retailer needs an AI assistant to start. Cleaning and structuring product data so it’s ready for AI is a good first step. It’s also worth considering what happens when someone lands on your site from an AI recommendation. Are they met with a relevant page or dumped onto a generic homepage? And as with paid search, you should track how those visits perform to build on what works.

AI agents are also moving quickly. Over half of UK retail leaders expect them to handle most customer interactions within five years. Walmart’s “super agents,” for example, manage orders, suggest meal plans, and even build recipes based on what’s in your fridge, all as part of a single conversation.

Impressive as it is, for those just starting out it makes more sense to walk before you run. My advice is to start small. Pick a simple, low-risk task such as back-in-stock alerts, see how people respond and build from there. And set limits. Left unchecked, agents can do unexpected things – from giving away freebies to forgetting who they work for. Think of them as brilliant new interns: full of potential, but not yet ready to be left unsupervised.

Owning Creativity in the Age of Platforms

Further upstream, where ideas are created, tested and refined, AI is transforming the pace of creative work. Amazon, Adobe, Meta and TikTok are all industrialising creativity, making it easier than ever for budgets to flow straight into their ecosystems.

The real opportunity is to harness that same capability in a way that serves your brand, not the platform. And we’re already seeing what that looks like in practice. Zalando now produces 70% of its campaign imagery using digital twins of real models, cutting production time from six weeks to four days and reducing costs by up to 90%. Channel 4’s AI tools help small businesses create scripts, visuals and voiceovers for TV ads without expensive agencies, while Sainsbury’s Nectar360 Pollen platform brings AI into the heart of retail media planning.

Sainsbury’s isn’t just plugging into what the platforms offer. It’s setting the rules for how those tools support its brand, its workflows and its data. And that’s the connection back to accountability. Testing at the pace you create only works if you’re also the one defining what success looks like.

Wiring the Grid

Most organisations I speak to already have AI pilots underway. The hard part is turning those experiments into everyday operations. That often starts with fixing the plumbing. Fragmented data kills good ideas. It’s one of those mantra’s that’s feels like it’s been around forever – generative AI means it’s no longer optional.

Take Tesco. It’s rolling out LiveRamp across grocery, clothing, mobile and retail media to pull Clubcard data into one place and power more than a dozen channels. The aim is to stop teams working from different dashboards and give everyone a shared view of what’s working. That’s the kind of plumbing that turns AI from a pilot into a habit.

It also means staying in control of how you activate and deploy AI. Platforms will increasingly own the pipes, but that doesn’t mean they should decide how you use them. And finally, it’s about rethinking the work itself. Curry’s, for example, uses “Action AI” to automatically generate store-level task lists, freeing up managers to focus on decisions that matter.

The next step (and the real leadership test) isn’t whether you’re experimenting with AI. It’s whether you still see it as a side project or whether you’re ready to treat it as the operating system of your business. If it’s the former, you’re basically using electricity to power a desk lamp – it works, but it’s hardly going to light up the neighbourhood. If it’s the latter, you’re wiring the grid and lighting up the market in the process.

« Back to All Posts