Startups and Silos: Purple AI's Open Approach to Leadership and Collaboration

I am Jake Sytner. I’m the CEO and co-founder of Purple AI.

Yeah, well, I guess I’ll start with the inspiration. The inspiration came out of Netflix. It’s a little-known fact. Netflix, when they moved on to streaming, essentially figured out that the data that they collected from adding metadata to their content and analyzing it allowed them to make their content. So House of Cards, the first in-house production, came out of a data-driven approach. They looked at which actors performed best in political thrillers, which directors and producers performed best in those categories, and they built the entire content, the entire TV show based on data. They performed well. And that inspired me to say, can we do this for another industry? And fashion is the clear choice here.

Sure. So the biggest two obstacles right now in fashion companies using data to plan which products to make Is essentially a two-sided problem. It’s a data problem in terms of data quality data management data quantity Data silos and more importantly, it’s a people problem To get actual fashion merchandisers and fashion designers to take a data-driven approach to their job Where they may not have traditionally done so is it’s an almost larger problem that is where purple AI comes in, we centralize all the data into the data hub. We include social media data, competitive data, all the first-party data, and trend data. We attribute that to our in-house technology. So breaking down all these products into the most fine-tuned rich data attributes and then putting this AI dashboard on top of it. So this AI dashboard acts as an AI copilot of sorts. So it allows them to do, just by planning this request, data analysis, data visualizations, product design, merchandising, really kind of bridging that gap between the art and the science there.

Yeah, so, you know, fashion is an interesting industry in the sense that, you know, there’s a bit of resistance, I would say, in most industries where, you know, there’s essentially a question of can art be enhanced by technology? You know, Netflix, as I said, proved otherwise with movies and films. And you know, fashion, again, there’s a bit of resistance, right? It’s been a traditional industry where people, you know, essentially use their gut instinct, maybe some historical data to decide what are we making, how much are we making, and what price are we selling it at. And what we’re doing when we come in here is giving these companies the confidence to say, okay, I’m going to work with this AI co-pilot to help enhance the art of fashion. And as we’ve seen, it’s gone pretty well. I can also speak to, I would say that the shining example of this in the industry right now is Shien. Shein is a fast fashion clothing company that came out of almost nowhere and is now doing huge numbers. And what Shein did better than anyone else is they essentially set up a data pipeline where they said let’s do data, you know massive data centralization aggregation enrichment analysis and just do almost completely data-based production. And as we see it’s doing extremely well for them. They’re staying on top of trends and delivering the products that customers want.

Fast Fashion in general, you know, Fast Fashion emerged essentially as this new model towards fashion. So you can kind of see the history of fashion through a lens of push and pull. So the push model was where fashion companies traditionally would sit in a room a year in advance or more, and they would decide, what are we making? And you’d have people with fashion backgrounds, fashion experts, who would be essentially weighing in and saying, we think next year, Block V would be great. Once those products were then designed, and manufactured. It was a heavy kind of push model, right? So a lot of marketing, a lot of advertising to try to get these products off the shelves and into consumers’ homes. Fast fashion was the first kind of model there where we said, instead of pushing the products that we made on consumers, what if we tried to pull the actual consumer demand, right? So Zara, H&M, these companies came in and said, we’re gonna just pull inspiration, pull actual data signals from consumers make products extremely quickly, and then deliver those to the consumers in enough time that they can fulfill that need. So the fast fashion model is becoming increasingly more attractive to all companies, not in the sense of cheap production and the manufacturing side of things, but pulling consumer demand through data signals and producing those products instead of taking a more risky guess on what consumers want. Also, I think last year 35% of all fashion products made were never sold, which is pretty astounding, it’s on top of the actual bottom-line effects there. Environmentally, fashion is not a great environmental sector right now. And reducing the amount of unsold products is a big win for the companies and also for the environment.

Yeah, so from a very kind of fundamental analysis there. Data analysts, data are expensive, hard to get, hard to onboard, and especially if you can’t afford them, it’s not gonna happen. The second level on top of that is just data, quantity, and quality, right? The name of the game in AI is data. I think a lot of people are focused on models, but the truth is you need hardware, you need talent, and you need data. And right now, if you’re a small team-size company, so purple coming in, taking care of all that and delivering that, you know, at a price, they can afford allows them to compete with the big companies who might have entire departments, you know, assigned to this.

Yeah. So there’s no data sharing within Purple. The way I like to describe Purple to companies is that our AI learns in the same way that TikTok’s algorithm learns, right? So it learns from usage, it learns from experience, but you can’t go into TikTok and say, Hey, show me what Stephen’s watching. And that’s what we’re doing here. Our AI is constantly learning to be a better merchant, to be a better designer, and to be a better employee for retail companies, but in no way is it revealing information to other companies. And so we know, you know, especially in fashion it’s a very competitive industry, it’s a very secret industry. And so data privacy is our top concern, but this AI model for all is I think the future.

Yeah, so I’d say the number one challenge right now is resistance, right? Again, fashion’s historically an industry with very proud people, you know, who are very art-focused. And the biggest resistance is to initially convince them, hey, you know, AI is not going to diminish the art that you’re doing, it’s going to enhance it. It’s going to allow you to make better decisions, faster decisions, you know, easier decisions, you know, less manual work in the process, but it’s not taking away the job of a fashion designer, it’s not taking away the job of fashion merchandiser. That human employee. So that’s number one. Number two, looking at larger companies, the data side is data management, data quality, and data quantity, these are very kind of intensive obstacles to overcome, right? So the first thing is always to go in and say, okay, let’s clean this up. Let’s index this, let’s vectorize this, let’s make this usable so that you can use this product seamlessly, have it access all information, and have that information be transferable.

Yeah, I can tell you my vision, which makes me excited, is the idea of an AI merchandiser who can truly work alongside humans. If you look at finance, for example, algorithmic merchandising is already big in finance. There are already companies who know that having an AI work as a trader already has huge upsides to those firms. I see a future where it has the same impact on fashion. A future where you’re not seeing 35% of products being produced and ever sold, an additional 30% of products being sold at essentially cost. But an industry where everyone who wants to can understand what to make, how to price it, how much to make, and to level that playing field in terms of competition.

Yeah, so, we like to have a very kind of open and I would say non siloed approach. So I like to hear from everyone. I like to allow everyone to have their domain, but also affects the work in other departments. You know, I see a lot of larger tech companies kind of have an issue where upward mobility is a bit of an issue, right? And being able to weigh in on large decisions is kind of reserved for the higher-ups. I think the advantage of being a startup is that anyone can get input, can give ideas, and give direction for innovation.

Yeah, I would say, and this is not going to be new, I would say the number one thing is, before you build something, you want to make sure that the thing has a need. And then the second thing, which I think is more of an AI-relevant application, is you want to make sure that the people who you want to use this product will use this product. So whether you’re doing AI for law AI for fashion or AI for politicians. Just cause there is a problem doesn’t necessarily mean that the people who have the problem will want to use the product. And then from there, again, the big three things for AI, hardware, talent, and data. I wouldn’t worry about the models. The models will at this rate continuously update, continuously be improved and replaced. Models can be replaced. Focus on what data can you access that nobody else can access to make your products completely, differentiated from the competition and to allow you to have a territory in this new AI landscape.

Sure. So you can always go to our website for more information. It’s purpleai.tech. You can always email me at Jake at purpleai.tech. And always happy to have a conversation about the industry, to meet new people, and to expand our networks. And excited to talk to anyone who wants to talk.

Thank you so much.