Latest / The Interline Podcast / AI VTO, Automation, Lawsuits, Personalisation, And Live Shopping
Transcript
- Grace: Welcome to the edit from the Interline, a new show where we run quickfire analysis on our pick of the most important fashion and beauty technology stories from the last seven days. I'm Grace, the social editor, and I'm joined by Ben, the editor-in-chief. But before we do that, I want to ask Ben to introduce the show, since this is our first episode. Ben: Hey, Chris, good to be here. Thanks for teeing this up. This will be, because this is the first one, it's going to be ever so slightly longer than our future editions. Just as an FYI, you're not going to have this intro from me every time. This is just to introduce a new format to our listeners. So anyone who listens to the Insaline podcast as it exists today is familiar with it as a 45 minute to 60 minute. kind of deep dive interview show where I pick a single guest and I talk to them about a very specific topic at length and in depth. That's great, those are going to continue, they are absolutely not going anywhere, but what I want to do with this and what you and I are aiming to accomplish week over week is to provide more of the Interline's perspective on the wider. ⁓ kind of the news and everything that's happening currently. ⁓ The interview shows are a great way for us to take a particular topic and really wrestle with it. These shows are intended to be much more of a finger in the air test, what's happening this week and what do we think it means? What are the implications from a fashion technology point of view? format here is also gonna be very different. So... Instead of me interviewing anybody, it's you interviewing me. So you're going to be asking me ⁓ questions about headlines and things that have surfaced from your research, our research over the course of the last seven days or so. the format for this show is that it's intended to be shorter, more digestible, and an easier listen. You're essentially dropping in on the kinds of conversations we have. at the interline regularly anyway. these shows are going to be more in the 20 to 22 to 25 minute range. We're going to get a little bit more specific about how that ⁓ over time. And I think we'll get a bit more refined on the format. But are intended to be shorter listens, easier listens, and more kind of expensive and broader listens. ⁓ So week you'll come to me with five or so ⁓ technology fashion and fashion technology stories from the last seven days. We'll talk about them. We have probably two, three minutes for each story that you and I will share and we will present our perspectives and then we'll just move on. ⁓ of these headlines come from our own research, from research we found elsewhere, from reporting we've seen elsewhere. We want to encourage people listening to the show to send in tips and ideas and headlines and things that they would like us to talk about as well to help close that loop. That's the very long-ish intro for this format, and you're not gonna have that in the future, so every week from now on, we'll get straight into it. And, ⁓ I'm gonna ask you to do that now. Hit me. Grace: The first question and the first story that I wanted to talk about was a story between DressX and Depop. So for those who don't know, DressX is a digital fashion platform, but it really got its start in COVID. now it's moving into more ⁓ virtual try-on. So did an event ⁓ and set up an AR try-on, AI-powered mirror. Users or visitors at the event could try on different digitized versions of Depop product then once they'd created a look they could actually get a link via email or text and then they would go on to Depop and they could buy the physical product. ⁓ found this story really interesting because it obviously showcases a more valuable use case for digital fashion. But that, Ben, I was really interested to know. Do you think this kind of AI powered virtual try-on experience going to help fashion resale? ⁓ beyond that, do you think this kind of ⁓ generated AI content ⁓ become a big thing in e commerce generally? Ben: So two things. have a bugbear about the term digital fashion. We'll put that to one side and we can pick that up at some point in the future. think what you're describing here is Deepop is a secondary ⁓ marketplace. And DressX has traditionally applied their AI virtual try-on to firsthand sale. ⁓ not ⁓ all, try-on been applied to ⁓ firsthand sale. Virtual try-on is having bit of a moment. ⁓ At the minute, we've previously referred to it as like fashion's kind of poison chalice from a technology point of view because so many people have tried VTO over the years with different technology underpinnings and hasn't seen broad adoption. It feels like AI is changing the game a little bit there for first-hand sale at least. ⁓ Now, secondary market has a kind of trust and experience delta ⁓ when comes to the way that it feels to interact with that versus interacting with the first-hand sale. Traditionally, if you buy something new, yes, you're getting a new product, you're getting a guarantee of quality and condition and experience and everything else from that. ⁓ new ⁓ experience is also where you typically have the cool experiences. That's where retailers and brands invest. So if you're going to get Virtual Try-On, you're usually going to get it in the first-hand market. So this is especially interesting because it brings it to the secondary market. And think if it closes that experience delta, so that people get a cool, fun... ⁓ engaging way of shopping for pre-loved items, it encourages more people to buy used. So that's encouraging from that point of view. ⁓ this one's really interesting because I think the standard of photography that people create when they're selling secondhand goods ⁓ a lot lower than the staged ⁓ and studio shot photography that you get for firsthand. Now, an AI VTO point of view, If you're ingesting really good quality, lay flat and mannequin and studio images and things from firsthand, then you have an easier uplift than you have more candid smartphone shots, less staged and less codified ⁓ ⁓ the secondhand market. Technically, really keen to see how this works. ⁓ Now, as I understand it, kind of, it's a go away, come back to you via message and email thing. I don't know what the time lag is in that. But for me, this is fascinating because it changes a little bit the way that people shop secondhand. And it's also an interesting technology problem that I really want to see how they solved it. Grace: So the next question is a bit more behind the scenes, and I want to ask you what on earth is going on with warehouse automation. I say that because I've seen two really conflicting stories this week. ⁓ first one was from ITV News, ⁓ it was talking about Asda's clothing line, and it's called George. And right now there are three warehouses for Asda's clothing line, but they're going to be reducing that to just one. And in that process, they're also going to be cutting a thousand jobs of people who work in these warehouses. ASDA reported that this is because they're introducing these new robots, so redline robots. On the flip side, I saw a very contradicting story from Amazon, are introducing their new AI-powered Proteus robots, ⁓ they're saying that these AI-powered robots can actually do the more manual work and Therefore it opens up new job opportunities for kind of higher skilled and more creative work. These are obviously two very contradicting stories, like I said. I'm just really interested to know what should we think about robots for automation. Are they going to create jobs? Are they going to reduce them? What's going on? Ben: Sure, I think specifically it's just, so you mentioned George at Asda saying they'll cut jobs by about a thousand. I think Amazon are saying that they will create something like 25,000. It's multiple tens of thousands of jobs from this. They're differing, ⁓ always with this thing, there's some nuance and there's some extra context. So for outside the UK, George by Asda is, ⁓ it's the world's most popular clothing brand. ⁓ And I think ⁓ you to read these kinds of stories. ⁓ with that in the back of your head, which is saying if you're consolidating a lot of the operational side of things, probably not because the brand is growing. then Amazon has precisely the opposite problem, which is that they have a huge amount of demand, a huge amount of volume, and a slightly different mandate when it comes to automation. So you can be automating and optimizing and consolidating something because it's of a fixed size, and you can go, okay, this is my clothing operations. I can see redundancies here. I can see opportunities here to do things this way. you can be saying, my business is growing. I have a much wider total addressable market. ⁓ and I need to lean on automation alongside hiring as a way of helping me to scale my business. Now, most brands in fashion and beauty are not Amazon. Most companies are not Amazon. Most companies are not forever growing and scaling and don't have the same kind of operational uplift and clout that Amazon has. I think automation in general in warehousing... more likely to eliminate jobs than it is to create them. think the creation is a temporary artifact of ⁓ growth here. ⁓ you look at groceries, so particularly here in the UK, if you look at a cardo ⁓ an example, a huge amount of their warehousing is fully ⁓ if that's a word, and ⁓ often when you talk about automation, whether in warehousing or in design or any other function, people will say, it's not going to eliminate jobs because somebody's job is to oversee the robots. it's become a very, very funny kind of fixed cliche, which is, we'll put robots in here, but it's fine because your job is now to maintain the robots. is not a one-to-one mapping of that. ⁓ And it's also fair to say that not warehouse worker, just as not every designer, not every patent engineer, not every supplier ⁓ coordinator whatever can or should be retrained to handle the automation side of things. speaking, I'm still a bit ⁓ radical in my middle age than I thought I would be in the sense that I lean more into the social side of this. ⁓ would expect me to be keener on technology automation. ⁓ If I was in this type of role, ⁓ would ⁓ expecting to be automated away rather than expecting to have more people working alongside me long term. Grace: The next noteworthy story I saw was about a lawsuit I saw in New York. a model called Francesca Pujos did a modeling job for a value-based retailer in Brooklyn called Rainbow USA. She did the modeling work, had a contract. Within their contract, it said that they could do some light editing work to the images. But then in March of this so 2026, ⁓ She noticed that the company, Rainbow USA, was using her likeness to generate new imagery. not only this, the images that they generated were actually very provocative. So just one example is that it showed the model's likeness kind of ⁓ on a stool spread legged, so a very provocative pose. ⁓ they changed her body, they changed her hair, but actually her face and likeness was the same. So it looked like she'd done this shoot. obviously has so many issues that we could talk about, but for me it stood out because it really feels like it's the tip of the iceberg of stories like this, ⁓ especially AI being such a new technology. ⁓ My to you, Ben, is why do you think a company, a retailer would think that this is okay? ⁓ And what can be done to avoid this kind of scenario in the future? Ben: Yeah, I love this story. ⁓ The original, a lot of people cited it from the New York Post article. I think we pulled the docket from the New York Supreme Court and went through it and specifically analyzed it. is potentially like a Bellwether case, or at least it's certainly one of the most publicized ⁓ in this space. We have an article being written by a former Model Now technology executive to appear in this year's AI report, which will be coming out in a few weeks time, which gets after precisely this issue. So I'm going to roll up some of the findings from the docket and some of his analysis in the answer that I give to you. I think what's especially interesting here is that this is not an AI lawsuit that is going after AI companies. I'm going to assume that there was a Gemini model behind this. So it was either one of the NanoBanana series or it was GPT image 1.5 or two. No one's going after OpenAI and no one's going after Google. ⁓ the in this case, Francesca, is specifically suing the retailer for the way that they are using AI. So that's ⁓ an distinction. This is not about how her face made it into training data. This is not about the fact that a model, unprompted and unbidden, Generated an image that looked like her. This is a company saying well, we've previously done a shoot with this person. Therefore we feel either kind of covertly or openly that we have a license to just remix her identity and her likeness. That's that's that's what that is the point that is at issue the specific statutes that she's going and she's going after are ⁓ kind of fairly typical, it's mainly around licensing. and existing structures that exist. And the defamation one is interesting, like taking somebody and putting them in a pose that they wouldn't have actually volunteered to do themselves. the fascinating part is that this is about behavior on the part of a brand or a retailer. And this is also behavior that a lot of brands will recognize that they probably do for internal purposes. So if you have a photo shoot that you already, you can already rely on, you have reliable models and you are coming up with a new collection and a generative workspace, it's going to be very tempting to just say, let's just bring in that model that we shot before as our reference. Let's use this to generate a bunch of new poses for this that we can use for internal selling or external selling afterwards. Right. That is not. how things actually work. That is not how traditional modeling contracts work. That is not how ⁓ likeness works. That's not how a whole bunch of things work. But it is very, very common practice. I think key part here ⁓ to rationalize for people to sit with ⁓ is the that it's very easy with the current set of generative workspaces to do work that is supposed to be creative and experimental and for in-house purposes. then to use the same workspaces and the same tools to generate marketing images, to generate e-commerce PDP content. ⁓ that's how lot of these workspaces are sold. They are sold on the basis that you do creative experimentation in here. ⁓ then you might as well generate your downstream facing assets as well. ⁓ you are not super clean about... who you are bringing in from a face and body and likeness point of view to this kind of thing. Before you know it, you slow walk into one of these lawsuits in the very near future. Grace: The next piece that came across my desk this week that I found really interesting was on the dark side of livestream shopping. was actually a vogue business piece and I believe it was written by Megan Doyle, who I think's written for the Interline in the past. Ben: Yes, we love Megan. She's written some great content for us before. Grace: Yeah, it was a really, really great piece. Megan goes into the of livestream shopping, ⁓ which was really big in places like China, now it's becoming really big in the West as well. And these are platforms like Whatnot and Tilt. ⁓ she talks about how popular they've become, ⁓ she also goes into the dark side of platforms, specifically how they employ these. ⁓ Manipul tactics to keep people really addicted and buying things on the applications. these are things like the fast-paced nature of how the streams work, also the gamification of them, and ⁓ of the fun novel bells and whistles ⁓ keep people really addicted to livestream shopping. And find this really interesting because as a fashion consumer, I can't imagine buying fashion in this way, but I do know that it is becoming very, very popular. I wanted to know from your take, Ben, do think that we should be concerned about live streaming? Do you think that it's here to stay or if it is a bit of a fad or a trend? Ben: Okay, so a couple of things. ⁓ Everybody should go read that story because ⁓ Megan's a clear thinker, a great writer. There's really good reporting in that. To take your second point first, live streaming is huge. ⁓ shopping is gigantic. I'm older than you. ⁓ So it's not something that I use if it's already not something that you use. However, I attended a fashion show at big studios here in Manchester a couple months ago. And as part of that, we had a big studio tour ⁓ and they a very big ⁓ kind of live stream setup that brands use, brands across ⁓ fashion, cosmetics. In cosmetics I believe they've been the host for ⁓ some of breaking live streams in terms of attendee figures in terms of like gross merchandise volume and things like that. If you're not aware of the scale of live stream shopping and that you don't interact with it, it can be hard to reconcile with just how huge it is. So I think this is a really timely piece from that point of view. is it going anywhere? No, just because I don't interact with it and you don't interact with it is absolutely the way that things are. would also put it ⁓ next ⁓ the kind of AI mediated experience that people are talking a lot about at the moment and the agent excited things because of body in two ways. ⁓ One is they're both tech enabled. They can't exist without it. You you can sell traditional ways without. technology. You can't do either of these things without a technology unlock. So all of the vast infrastructure behind live real-time video streaming and interaction is behind the live stream shopping phenomenon and the development of large language models is behind the AI shopping phenomenon. They're both unlocked by that thing. They're both also potentially pretty problematic from a experience point of view because as Megan points out in that piece, streaming creates a sense of urgency and scarcity that encourages people to make purchases that they wouldn't do otherwise. ⁓ you look at it as just QVC for younger people on the internet, you're treating it as more benign than I think it actually is. ⁓ what you have ⁓ ⁓ are used to instant gratification ⁓ these kinds of platforms. And gratification lends itself to impulse purchasing. If you pair it, as a lot of platforms do with buy now pay later models, you are creating a flywheel for instant debt generation as well for younger people. same is true, and we've written about it before for AI. ⁓ a very short hop from personalized shopping interactions through chat GPT or Claude or Gemini ⁓ prescriptive behaviors. AI chatbots are very much optimized for engagement ⁓ for giving people what they want. It's not hard to see how that becomes kind of incentivizing purchases and forcing them people's hand there. Again, pair those with Klarna and so on. And before you know it, what we have is a system that really pushes people towards purchase a way that I suspect might come under regulatory scrutiny ⁓ the long run. But ⁓ stream shopping, huge, absolutely gigantic. Grace: The final story is something that I saw on the Amazon blog this week, and it's to do with AI-generated designed merch. think there's been a time in everyone's life where they've wanted to create some kind of customized merch, whether that's a t-shirt, a hoodie, or a water bottle, something like this. ⁓ Amazon's Alexa for shopping is actually making this even easier ⁓ any design that anyone thinks up. ⁓ They can now generate that design using AI and put it on a t-shirt or a hoodie, like I said. not only this, but Amazon is now able to create these products on demand. And believe they're also using their prime eligibility to also ship these products. On one it goes without saying that this kind of service could be so popular. You can see ⁓ why it take off so much. ⁓ On the hand, people are really not liking AI at the moment, especially young people. So you could also see that people would really want to rebel against this kind of service and not use it. What is your take Ben? Do you think this is going to be really popular? Do you think it's going to take off? Do you think it's a good thing or not? Just what do you think about it? Ben: So I won't pretend to have a good read on how younger people feel about AI in particular, except what is out there as published data, which is not encouraging. AI is anathema to what a lot of younger people want in their creative brand expressions and in the way that they are marketed to and story told and engaged with. the other hand, I also think there's a whole lot of people who are not artists, don't want to go through the hassle of becoming artists, ⁓ do want to creatively express themselves in what they were. ⁓ This is something that traditionally you would have... had to go through kind of screen printing with minimum order quantities and stuff to do. If you want to make your own t-shirts, you have to find supplier of high quality blanks. You have to do the artwork yourself. You have to prep the artwork to be ready for traditional methods. Now, Amazon one of the largest customers of Kornit specifically, but I suspect maybe other ⁓ kind of digital direct to garment printing technology hardware. They've been buying that up for a while and they use it in their private label. ⁓ white label ⁓ kind of product lines for t-shirts and hoodies and things and so on. So they already have the infrastructure to do this. That is not new. You can do unit of one high quality digital print onto ⁓ blanks any quality, ⁓ right? there are whole operations aside from Amazon ⁓ set around the UK, the EU, the that do this kind of thing. They do effectively drop shipping ⁓ where they will also print the artwork onto the t-shirt for you. Now, As for whether people will actually use this, I suspect that they might because saw last week that Amazon were adding kind of image generation to their search ⁓ if somebody was saying, okay, I want a blouse that looks roughly like this, this is the kind of color that I want, this is the kind of pattern that I want, it generate fake images of fake products and then afterwards it would try and map those to ⁓ real ⁓ Guess a really good ⁓ way ending that transaction instead of mapping it to some third party? products that you stock as a marketplace, just encourage people to print their own. somebody's searching and they're saying, I want a red t-shirt and I want this, ⁓ I want this it, I wanted this kind of pattern. ⁓ It doesn't have be artwork, it can be pattern print. ⁓ Just let them it themselves. ⁓ This is a example, sorry for the pun, but this is a prime example of ⁓ where you build sufficient infrastructure for something. Digital printing in this case paired with... a mixture of in-house models and your own Amazon Bedrock hosted third party and open source models, all of that on and all of a sudden you can offer a very, very different model of clothing. ⁓ don't know how this will play out because I said I don't have an exact read on how people feel about it, ⁓ it will work from a technology point of view, it will work. And that's interesting. Grace: Amazing. Ben: Fabulous. That's the end. That is the end of our first installment of this. Grace, thank you for quizzing me. always, I'm looking forward to doing this again next week. And you will see these shows now come out on a Tuesday. So every Tuesday, you'll get this new edit in your inbox if you're a subscriber to our mailing list. or you'll see it pop up on Apple Podcasts or Spotify if you follow us there as well. weekly deep dive interview shows are moving to Thursdays, so if you're here for those as well, you will get those from me. We have a big program of those lined up over the next ⁓ four months. We're booking guests out too at the moment. And there's some fascinating conversations that I'm looking forward to having there. So that side of things won't change. If you want to see me interview people in a deep way, come here on Thursdays. ⁓ If you want more of this, if you want more kind of broad industry news, my perspective, Grace's perspective on what's happening week on week in a more condensed format, we're going to keep doing this as well. And we look forward to welcoming you back every week.