AI-Native Brands Are Inevitable, But Maybe Not In The Way You Think

A lot of virtual ink has been spilled about the potential for generative AI to “democratise creativity”. Individually, of course, you’re entitled to feel any way you like about that idea; The Interline can, charitably, see both sides, even if the result looks, to us, like a race to the bottom either way. 

But it remains true that if someone has a vision that they previously struggled to bring to life, either through limitations in tooling or a lack of opportunity for skills development, then image and video generation does afford them an absolutely turnkey way to get that vision out of their heads and into the world. They still need to find fairly traditional ways to then translate that vision into producible product, of course, but AI has blown a lot of the initial friction away, for better or worse.

It’s also logical to assume that, of that cohort of idea-havers who were held back by the need to learn systems and put talent into practice, at least some will have good ideas – concepts that are not just idiosyncratic and individual, but that are innovative and have genuine market fit.

When people talk about “AI native” brands, this is generally what they mean: a cohort of people who either haven’t been able to crack the existing fashion system because their creativity has been trapped behind a wall, or people who see an opportunity to come from outside the industry now that the thing that defines it in the early stages (design) is now open season for everyone.

And make no mistake: there will be plenty of these brands and designers over the next few years. The levelling (or compressing, depending on your perspective) effect of generative AI is going to open new doors to fresh competition across the full spectrum of quality and viability.

But by and large, generative image models (or even generative pattern development tools) are, by themselves, going to provide a new pathway for people who want to try and run the existing model of fashion better – or at least with shortcuts. Lowering the barrier to conceive new concepts, or to iterate on existing ones, might increase efficiency and throughput in the early parts of the product journey (and then compound problems in adoption a little further on, but we digress) but it won’t automatically unlock a fundamentally different route to market.

And this is more than just a feeling. The idea that deploying AI will increase the speed of existing processes but fall short of the more comprehensive transformation it promises is being validated across other sectors this week.

The most noteworthy story in that bracket is this one: that a high-profile enterprise deployment of AI (specifically NVIDIA’s suite of software and hardware) at Bank Of America had stalled, because project heads and end users felt they had been sold “a Formula 1 race car” that they were unable to effectively drive, to stretch a metaphor. This sentiment was aired through leaked emails from NVIDIA’s sales teams, as was their response to it, which will sound very familiar to anyone who’s ever worked in software sales: the customer lacked the skills in-house to make effective use of the technology, and the vendor would need to “collaborate on deployment”.

Software customisation is not necessarily a terrible thing – certainly not in the early stages of any new technology adoption cycle – but this story has a bit of a darker heart. What if Bank Of America pays for a massive chunk of implementation service days, or hires in the machine learning operations talent NVIDIA seemingly believes it needs, only for the target benefits to still not emerge?

The answer, in that scenario, might just be that AI could be capped at delivering iterative improvements unless it’s employed by entirely new companies, or unless existing companies are rearchitected in some pretty profound ways.

That latter point is one that also earned some airtime this week following a Business Insider roundtable held at the World Economic Forum, in Davos, last week. At that event, which was explicitly convened to address the questions of why AI has had so little measurable impact on work, Deloitte’s Global Chief People & Purpose Officer had this to say: “84% of work processes have been left in their legacy state when adopting AI and have not been redesigned”.

Anchored to that statistic is the realisation that the vast majority (80%!) of the ways that companies currently operate might need to be at the very least refreshed, or more likely rebuilt.

And while there’s no evidence at this very early stage to suggest that OpenAI’s newly-announced partnership with PVH – publicised on Wednesday – will fall into that same trap, the scope of that major initiative definitely sounds progressive, but it also sounds relatively safe rather than disruptive, targeted to operate in an efficiency / productivity capacity in very familiar lanes like “early-stage product creation […] demand planning, inventory optimization, and consumer engagement.”

All of which suggests that AI-native brands – i.e. companies that don’t have the same process legacy or technical debt of competitors who’ve been operating much longer – should have a significant advantage. If incumbents need to redesign more than 80% of their processes to bring the touted transformative benefits of AI into reach, then new entrants are being given a potentially potent wedge.

But The Interline is increasingly convinced that being “AI native” is about much more than using generative AI to explode the beginning and end of the product funnel, or even to improve throughput in the middle. 

And while our money would be on the rise of AI native brands, they might not be disruptive in the ways that people expect.

In fact, by definition, they’re going to be disruptive in the ways we don’t.

Consider the companies that have defined the last couple of iterations of the web, and that, crucially, couldn’t have existed without those step-changes in the communications infrastructure that connects us all. 

Famously, Facebook and other social platforms came about as much thanks to the open source “LAMP stack” (Linux, Apache, MySQL, PHP) as they did the vision and verve of their founders. These were businesses that could not have been built previously, and businesses that relied on some measure of democratisation of previously skill and capex-gated technology to be born.

And the next stage of the web? eCommerce combined with social was, inarguably, the reason that a massive wave of direct to consumer (DTC) brands were able to start, and low-rate capital injection was how they were able to scale. 

The fortunes of those companies afterwards has been extremely divergent – Allbirds is still truckin’ and social / eComm-native brand Vuori is now worth in excess of $5 billion, while high profile DTC casualties like Peloton have lost the majority of their value – but the one thing they have in common is that all of them were either born or forged in the transition from the static web to the social web, and then empowered by the rise of the now-codified eCommerce stack, across storefronts, payments, fulfilment, marketing, and much more.. 

These were not companies that could have pre-dated their era of web technology any more than the upcoming wave of genuinely AI-native brands would have been able to exist without theirs. And while it’s easy for analysts to look back and point out that all the conditions for the DTC explosion (and the contraction afterwards) were there, it took the launch of brands who built with those conditions as their foundations for the realisation that a different type of business had been born to set in.

So what does a true AI-native brand look like? We can take a guess (and we probably will, in this year’s AI Report, coming this spring!) but the essential thing to remember is that these kinds of step changes feel inevitable in retrospect, but are difficult to predict the shape of in advance.

Perhaps not coincidentally, we’re now a decade or so past the launch of most of the eComm-native direct-to-consumer brand stories, and we’re at a moment where consumer confidence – at least in the USA – is at a twelve-year low. Those sound like precisely the kind of market conditions for a fundamentally new way of making and selling fashion, with AI not as a tool but as the foundational tech layer and the primary route to market.