Only 12% of companies are getting real results from AI today. That figure opened Emerce EDAY 2026 on 5 November at the Kromhouthal in Amsterdam. It set the tone for the rest of the day: less hype, more questions about what AI actually delivers in practice.
The Tilroy team attended sessions on product data, app development and design systems. Some speakers were hopeful, others confronting. Together they paint a clear picture of where retail is heading.
Below we share six takeaways, each translated into what it means for retailers.
The opening keynote came from Irina von Rosén, investor and former tech executive who guides companies through AI implementations. According to the PwC 2026 Global CEO Survey, only 12% of companies saw AI both lower their costs and raise their revenue. More than half saw no change on either.
Three causes kept coming back:
Her advice was simple: clean up the mess first. For retailers, that means making sure stock, product data and customer data are correct in one place before adding AI.
A PIM software vendor showed how product discovery is shifting from Google to AI assistants. Customers no longer type "espresso machine". They ask for an espresso machine that fits a small city apartment. If the dimensions or use case aren't in your product data, you won't appear in the answer. Your product ends up on the "invisible shelf": it exists, but nobody finds it.
According to the speaker, good product data for AI rests on four pillars:
The Leatherman example stuck with us. Reviews revealed that electricians use the multitool to strip cables. The brand discovered a new audience purely from what customers wrote about the product online.
A session on a large app study (almost 14,000 users, 400 apps in the Netherlands, Denmark and Sweden) showed that most apps have their basic features in order. Those who want to stand out do so with innovation and personalisation: the "intelligence layer".
That layer comes in three forms:
The further along this scale, the less visible and the more valuable. The lesson for retail: start from the friction your customer experiences, not from the question of where you can add AI.
A design system is more than a style guide. It's a central set of components, code and documentation. Now that coding agents build interfaces, the audience for that documentation is shifting from people to AI.
Through MCP, an AI assistant can consult the entire design system directly. The better it's described when to use which component, the more consistent the result. The agency behind the session saw the accuracy of generated components rise from 80% to 98%.
Two modes proved useful: a strict mode for production (a form should always look the same) and a looser mode for concepts (a new landing page may vary).
A session on "curiosity infrastructure" started with a game. Who has sat in a meeting where the decision had already been made? Who stayed quiet because they didn't know the answer? After six questions, most people in the room had folded their hand into a fist.
The message: now that AI makes answers abundant, those who ask better questions win. That takes four things: permission not to know, safety to share ideas, structured experimentation and an honest story after a failure. Three questions to start with tomorrow:
Pau Aleikum Garcia of Barcelona-based Domestic Data Streamers showed what happens when you don't use AI for speed and control. His studio turned the idea of camera surveillance around and built a model that detects when politicians fall asleep in parliament.
The most moving project reconstructed memories of people in refugee camps using image generation. The image didn't have to be pixel-perfect. What mattered was that the memory was right, so the story could be passed on.
For an audience of marketers and developers, it was a useful reminder. The same technology we use for conversion can also bring people closer together.
The common thread of the day: AI rewards those who have their foundations in order. That goes for product data, for processes and for the culture in your team. Three things retailers can do today:
At Tilroy we build retail software that brings stock, product data and customer data together across all channels. That lays the foundation AI can build on. Discover how Tilroy works.