The Untapped Potential Of Upstream AI

This article was originally published in The Interline’s AI Report 2025. To read other opinion pieces, exclusive editorials, and detailed profiles and interviews with key vendors, download the full AI Report 2025 completely free of charge and ungated.

---

We’re in the era of AI everything. AI customer service, AI website creation, AI designed beer (not a joke). In just a few years, it’s become investors’ dream, a cultural meme, and quite literally everything in between. And though its sudden ubiquity has made some eyes roll – it’s clearly in the process of evolving every industry under the sun. And the fashion world is no exception. 

But what does this reshaping look like, really? AI’s ability to organize information and provide answers has delivered unprecedented efficiency in customer service and logistics. And its capacity for image generation has certainly created marketing content and inspired designers’ collections. 

But these uses, while impressive and at times profitable, come with backlash. Not simply because of their tendency to consolidate value and replace workers, but because of AI’s massive consumption of energy and water. 

The fashion industry already has a sustainability problem. For the last two decades, it’s been increasingly dominated by stats like “2700 litres of water per t-shirt” and “Between 10-40% over production”.  The apparel industry creates over 100 billion garments annually—far more than the global population needs—driven by fast fashion and relentless consumer demand. This overproduction and overconsumption strain natural resources, generate waste, and contribute significantly to global carbon emissions. 

The industry also knows this. Since the explosive 2018 IPCC (International Panel on Climate Change) Report that outlined our short timeline to address the climate crisis, most global brands have made ambitious promises to reach set zero by 2040. And while some brands have made progress in addressing their direct emissions, “scope 3 emissions” (those created throughout their supply chain) have remained very difficult to address. 

This is because the most impactful part of the industry is in fact, the products we make. So if this amazing tech is really going to fundamentally reshape fashion for the better, it needs to address challenges further upstream. 

Luckily, that’s what’s happening now. Thanks to recent advancements in the field of computational biology, it can. New forms of AI have the ability to affect not just how we sell clothes, but what those clothes are made of. 

And it all starts with plants. 

Cotton, the most sought-after natural fiber for the fashion world, is made of nothing but the long cellulose sheaths connected to the seeds of the cotton plant. And the varying attributes of these fibers (their quality, texture, color, carbon footprint and so on) are largely determined by the genetics of the cotton plant. 

New forms of AI have the ability to affect not just how we sell clothes, but what those clothes are made of.

And it all starts with plants.

Humanity has been selecting beneficial plant characteristics to improve agricultural traits since humanity saved and replanted seeds from the best performing plants in the field at the dawn of agriculture, but that process is slow, expensive–with current technology–and full of guess work. That’s because the genome is an incredibly complex library of dense biological data that has been historically very difficult to decipher. Luckily, analyzing and making sense of massive data sets is precisely what AI does. 

Today, biologists are able to leverage machine learning to increase the efficiency of the evolutionary process, creating more resilient and sustainable crop varieties that can address industry challenges from the very beginning. 

This is done by training an AI model on the rich data within the genome of a crop species, like cotton. The AI studies the genetics of hundreds of domesticated and wild varieties (remember that in nature, the more diversity the better) in order to understand how that plant genome works, and more importantly, what it’s capable of.

Put another way: Traditional Breeding works more like “guessing” – where breeders cross plants based on what they can see, and then hope for a desired result. AI breeding works more like “asking” – where scientists can ask  “which plants should I cross to get these traits” and the AI can tell them.  

Using AI to select which plants to breed is a totally new way to approach crop development, and has massive implications for the entire value chain (not to mention, the planet.)

First – it drastically reduces the amount of time it takes to bring improved crop varieties to market. Traditionally, it takes the industry 8-12 years to introduce a new variety. AI can do this in ½ the time. And with global agriculture facing pressure from a changing climate and growing population – the pace of innovation is huge. Our solutions have to outpace our problems, and by increasing the speed of evolution we can create much needed resilience into our food and fiber systems. 

Second –  new varieties can be produced at a fraction of the cost of traditional breeding programs. And less cost of innovation means…more innovation! So rather than developing one cotton variety for multiple markets, breeders can develop different varieties that are evolved for the unique challenges of being hyper-specific. For example, our new Avalo Cotton variety is specifically evolved for the increasingly hot, arid conditions in the Texas Plains – allowing it to thrive without any added irrigation and far less fertilizer. Regionally adapted varieties also make it possible to shorten supply chains by making more plants possible in more places. 

Finally – it allows for “multi-modal” trait development. Most breeding programs focus on developing one trait at a time. But because the AI allows breeders to look at the whole genome (and all the interactions between genes) we can select for more complex concepts like “profitability” which includes balancing attributes like yield, inputs, fiber quality and more. 

What AI is unlocking isn’t just better breeding—it’s better sourcing.

This ability to shape multiple traits at once is where this approach will be able to reshape the fashion world. It means being able to design crops with the whole supply chain in mind. New cotton varieties deliver high fiber quality with fewer inputs (win for farmers) and create up to a 30%* reduction in greenhouse gas emissions (win for brands, people and planet). And this ability to include farmers, gins, spinners, mills and producer needs into the “sustainable innovation” process delivers the value needed to stimulate mass adoption. 

What AI is unlocking isn’t just better breeding—it’s better sourcing. By harnessing the full potential of cotton genetics, breeders can develop products that simultaneously reduce environmental impact, increase efficiency and improve performance, delivering value at each point along the supply chain. Imagine raw materials bred to thrive with fewer inputs, deliver longer staple lengths to meet spinner and mill requirements, and maybe even have natural pigmentation – all coming straight from the field. A Win/Win/Win that addresses Scope 3 emissions, water usage, procurement risk, performance requirements and consumer demands all at once. 

The best part of this all, is that this application of AI is amplifying, not replacing, people within the supply chain. It’s creating and distributing value, rather than consolidating it. And this is critical at a time when farmers and other upstream actors are struggling to get by. It’s easy to forget that fashion is intrinsically linked with agriculture, so “industry reshaping” has to include them. 

Now that it can? Well, that’s AI worth the hype. 

*Based on 2024 estimates. 2025 LCA currently in progress.