While AI might be what comes next for retail, at NEXT successful adoption lies as much in what the firm isn't doing

I know that there's going to be a groan when Chief Executives start talking about AI, so I'm going to apologize in advance.

A perceptive assessment from Simon Wolfson, Chief Executive of UK fashion retailer NEXT, but one that doesn’t prevent him from outlining his firm’s own approach to the technology du jour.  It won’t come as a surprise to anyone, but his pitch starts from the premise that NEXT is well-placed to benefit from the shift to an AI-centric world:

We've modernized pretty much all of our software platforms, with the exception of Finance, over the last six years. We have transformed the way the company handles data, we've made sure that it's sort of universally-accessible across the group, that it's consistent across the group. So the combination of modern software and high-quality data means that we are well placed to adopt AI.

Certainly that sounds like the recipe for AI success that vendors have been more pragmatically pitching of late - get your data foundations in place, then build AI on top - so how far down the AI journey is NEXT at this point? Wolfson says:

The degree of adoption that we're getting across the business is very different. The areas that have adopted it the most aggressively are technology, contact centers and e-commerce. Use of AI to envisage product and forecasting has made some progress. Where we have invested, we are definitely seeing AI resulting in productivity gains that is translating directly into not just better software and better service in the call centers, but also lower costs.

Interestingly, when outlining NEXT’s strategy for AI, Wolfson argues that the most interesting angle lies with what the firm is not doing:

We don't have a central AI department or a Chief AI Officer, a CAIO, I think it would be called. (It sounds like ‘goodbye’ in Italian!) We don't have that, and the reason we don't have that is because the nature of what we do across the different functions of the business is so different that to have one department trying to service all of those would be extremely unproductive. 

Ultimately, the value of AI is not in the technology, it's in the applications it supports, and the design of those applications. What it does for the business can only be understood by the people who are running the business, not people who have access to the technology.  

The best comparison I've heard is that it would be like having a central Spreadsheet Department and a Chief Spreadsheet Officer. Spreadsheets are used by everyone in very different ways across the business. Same with AI. It's got to be application led.

Collaboration

That doesn’t mean that it’s an AI free-for-all, he adds:

That's not to say we're doing nothing centrally. We do talk a lot about it. We encourage people and help them. Our IT department provides access to technologies. It ensures that the technologies we use are secure, most importantly, and it also monitors the cost of the AI tools that various departments are adopting.  But we haven't adopted a one-size-fits-all approach to this. Each [business] director is very much going to have to be their own Chief AI Officer if they want their part of the business to succeed.

Don’t under-estimate the power of collaboration here, he urges, noting that the way NEXT works helps here: 

 Pretty much all of our directors see each other every week at our trading meeting, at least once a week, if not more. And the directors who are most advanced [in AI adoption] are actively working not just in their department, but to help the other departments and show them the sort of things that AI can do for them and share people and technology providers with them.

But there’s still a long way to go, Wolfson admits:


Even in the areas that we are the furthest ahead, my guess is that we're scratching the surface. There is so much more to go at, not just in terms of cost effectiveness, but also productivity in terms of what people can do. So I think this is a huge opportunity.

Warehousing is an example of a business area that hasn’t got a notable AI footprint, he argues, but this isn’t a case of attributing blame:

I'm not singling out warehouses because they've done anything wrong or because AI isn't applicable to warehousing. Actually, I think AI could be brilliant to warehousing in terms of handling all the operational management of the warehouse, re-forecasting, scenario planning, optimization. [There are] lots of variables that AI could really turbocharge the management of our warehouses. But they haven't had time quite rightly to worry about AI because the thing they're most worried about is ensuring that we've got the capacity in warehousing that we need to grow.

Dis-intermediation danger?

Of late the major Large Language Model (LLM) providers have made in-roads into the retail sector on the back of the likes of agentic commerce. No retail event seems complete without the signing of (another!) exclusive partnership with the likes of OpenAI. But this does raise the question of how much of a risk of dis-intermediation such alliances pose to the traditional retail operating model.

It's a big question, admits Wolfson, but he adds:

The first thing to say is rightly or wrongly, it's not something that we're overly concerned about at the moment. I think the dis-intermediation that we're talking about would be the dis-intermediation of the website, rather than any other cost at the moment. [AI LLM providers] could switch just a fraction of their data centers into beautiful clothing warehouses and ChatGPT would have the infrastructure, but at the moment, they don’t so you're really talking about the dis-intermediation of the shopping bag and the selection process.


There are economic and operational problems with that thesis that are yet to be solved, Wolfson contends:

The economic problem is that if you look at our average order value, net of returns, which is around £70, the cost of delivery to the consumer is around £5, about six percent. Wherever an intermediate website goes, it’s got to go to a number of retailers. If it comes to us, then fine, it's just another form of advertising! If it goes to more than one retailer, let's say it goes to four retailers rather than one, you end up with that six percent being multiplied by four. So the economics are, someone, somewhere, has got to pay for that additional 18% of cost that you'll get from splitting the order across multiple websites.

The operational problem is even more of a challenge and applies specifically to the clothing retail category:

That is how you handle returns because if all of your online order goes to John Lewis or to Marks & Spencer or to NEXT or to Very, you can take the whole order back to any one of their shops, scan the items, and you're credited instantly in the way in which you paid. For an intermediary to do that, you've got to know where to take the item back to. Which of the sub-vendors do I take it back to? How do I return it to them? And then how do they communicate with the intermediary, that the intermediary has got to repay me?  So there are big customer service issues with it.

There’s a familiarity about all this, he suggests:

At the moment, it feels to me very much like what people were saying about marketplaces versus stocked retailers because the real asset in trading online clothing is the logistics infrastructure and the product, not the website. So I think that is a direction [AI vendors are] unlikely to go in. If you look at the difference in Google’s approach and ChatGPT’s approach, Google is still taking the approach of passing the consumer straight through from their AI engine to one or other retailer, so it becomes an enhanced form of advertising. I think to that extent, it's very exciting. Basically, the better search engines can find what customers are looking for, the more they'll buy, which has got to be good for the industry overall.

But things could change, he concedes:

What is really interesting when we stand back from what we do day-to-day, and I think about the sorts of things that cross desk and my colleagues' desks in terms of new AI-driven marketing technologies with Google, all the things that are driving the business are completely different to the sort of things I was doing 25 years ago. Twenty-five years ago, it was still very exciting, but it was about stores and, believe it or not, catalogs. Getting more catalogs printed and printing those catalogs was central. 

So what the business does today is unrecognizably different from what it was doing 25 years ago. The thing that really has not changed at all are the principles upon which that growth and the ethos of the business...First of all, absolutely everything we do, we have to, hand on heart, believe that it is giving good value to our customers.

And there are two important lessons overall:

One is, in the good years, don't get cocky. And in the bad years, don't go bust!

My take

No groaning here about what looks like a pragmatic retail AI worldview that hasn’t been caught unawares by the hype cycle.