My 22-year-old son does his clothes shopping on TikTok. So how does a brand like JD Sports Fashion reach him, squarely in the middle of its target 16-24 demographic, to buy its products there? After a two-year transformation of its online presence and with $15 billion in global revenue at stake, JD Sports now has an answer to that question. Antonia Hansen, Group Director of Product, took to the stage at the MACH X conference in Amsterdam this week to discuss how its new platform will help it adapt to the future of agentic commerce.
But before we get into the detail, let’s jump to the point in her presentation that leapt out at me. This is not a story about simply adding TikTok or whatever else as a new checkout channel into the omnichannel mix. It’s about a fundamental change to how brands present themselves to consumers in an agentic world — about recreating the brand experience in social media settings its <a href="https://bitcomme.com/american-waters-customer-care-agents/” title=”American Water’s Customer Care agents”>customers are engaged with, just as it does today in its stores (where if you’re not in that demographic, you probably hate the loud music!) She explains:
Our customers… don’t want to just buy a pair of trainers because they need trainers. To be honest, they probably don’t need another pair of trainers. But they’re inspired because they want that extra pair of trainers. They want that latest Air Max 95, or whatever the latest shoe is, and that’s where we want to drive to as JD.
At the moment, our digital platforms are relatively transactional. To be honest, you pick up the JD app when you want to buy a pair of shoes or a hoodie. Where we want to get to, and where we’re able to now build towards with our new composable stack, is being much more of an inspirational ecosystem…
Personalization will become really key, and making sure that we have a continuous feed, so that our customers are picking up our digital experiences when they want to hang out in the world of sports fashion, as opposed to just buy something.
In this agentic future, the transaction will have to become as seamless as possible, rather than a separate action that the consumer has to go and do. Getting there will be a multi-step journey, with plenty of iteration along the way as the technology evolves and consumer behavior responds. For now, the focus at JD Sports is on making existing channels as rich as possible with AI interactions.
Agents completing purchases
Having started to go live with its new platform in August, the first step has been to build and roll out a one-page checkout, replacing the previous multi-page process. Meanwhile the company is testing new purchasing routes, such as Gemini Instant Shopping, which means that shoppers in North America can now search for a product, click on the JD Sports search result, and immediately purchase with Google Pay. Another US initiative is native in-browser shopping for certain brands, where LLMs get JD Sports’ product information directly from its product feeds rather than having to scrape the website. Looking ahead, the goal is to be ready for the day when agents are completing purchases on behalf of consumers. She explains:
In one to three years, we then see that starting to grow and grow within the standalone AI assistant channels, be that Gemini, be that ChatGPT, be that Claude. Then over the long term, what we see is, actually once we really nail that agentic purchase moment, we have made it way easier to shop.
Where we see that really looking to grow is a shoppable everything future, where suddenly that really complex retail transaction that I talked about is simplified, and agents can do it. There should be no limit between what’s retail and really what’s media, and then what you can shop to.
None of this would be on the JD roadmap without that initial change to a new platform, working with MACH-compliant vendors such as Algolia, Amplience, Bloomreach, commercetools and Stripe. It replaces a highly customized, homegrown monolithic stack, where it had become so difficult to make even the smallest change that there had been no major releases in more than twelve months. That changed as soon as the roll-out of the new composable stack started in the brand’s major markets in August, with three major features already released, and four more planned before the peak shopping season ramps up in November. Now the company feels able to face the agentic future, says Hansen:
Agentic commerce is a huge part of what we can now enable with our composable stack. And again, if this dawn had arrived two years ago, it would have felt so scary because we would have been so far away from having a stack that can really be built upon at speed.
One crucial difference is that the ability to build new features quickly also allows the team to fail fast. She comments:
When you’ve spent two years building something, it’s very expensive just to throw it away. What we now have with our new technology stack is the ability to build fast. So it becomes much, much easier to go, either, ‘Actually, that wasn’t a good idea,’ or, ‘It’s not the right idea for right now.’ That’s been really interesting within our agentic commerce experimentation as well.
Breadth of content
For example, a feature that allows a customer to virtually ‘try on’ an item of clothing by uploading a selfie worked well, but it wasn’t making a noticeable impact on sales, which didn’t justify the inference costs of running it. This was therefore paused. Another new feature that did pass muster was a conversational agent on the JD website that customers started using in unexpected ways. Hansen explains:
What our customers wanted was more information. They wanted to know, how thick is this material? How waterproof is this material? How does this fit? Which was a really interesting learning for us…
We realized there’s actually so much more that our customers want from our product description, for our product display page. I think something that’s really powerful about conversational and LLM-type functionality is that you get so much more of an insight into your customers. Before, you have to just interpret where they click, or interpret where their focus is. Now, through these sorts of features, you can genuinely understand what they’re trying to do in their own words, which is massively powerful and has been hugely helpful for us in terms of understanding what features we want to drive next.
One consequence of this finding is that, although on one hand, AI has sped up the process of content creation — so that, for example, product and marketing images can now be produced so much more easily — on the other hand it has increased the breadth of content that needs to be created. And that content not only has to be produced for humans but also to satisfy the demand coming from agents. She explains:
Whereas before, a human [is] probably not reading more than 10 blue links for a discovery journey, AI can just easily index hundreds and hundreds of links. So the need for quality content, and structured content, gets so much greater.
We’ve recently expanded a number of different teams to include roles specifically around AI discoverability and GEO, to make sure that we’re structuring our data in the right way, that we’re providing the right level of data, and in particular for us, as a multi-brand retailer, injecting our proprietary understanding of the product with as much richness as possible.
My take
One theme that kept on recurring during this week’s MACH X event was the need to add more content to answer the questions that are coming in from agents. When buyers are able to ask questions that go outside of the boundaries set by deterministic systems, brands get much more information about what really influences their purchase decisions — and much of that new learning is unexpected. I’ll have more from other sessions at MACH X in the coming days.
But Hansen’s presentation about the changes that JD Sports is preparing for was even more of an eye-opener. It conjures up a future for retail that lifts it even further beyond the domain of the brick-and-mortar store, and more intimately into the lived experience of consumers. I feel that calling it agentic commerce is almost a misnomer, because that phrase focuses all the attention on the technology. The big changes will come from how the tech augments human engagement rather than replacing it, and the cultural impact of that.
