Calculating The Cost Of A Prompt

When you’re building a new company in the AI era, it seems like it would be second nature to introduce Generative AI workflows. We hear so much about “AI native” businesses being inevitable that we’ve all but forgotten that some companies might want to choose a different path, and that plenty more find themselves torn over whether the possibilities of AI are worth the potential trade-offs.

It’s not just entrenched companies that think this way, either. We’re building a forward-thinking, tech-first, data-native fashion brand, and we hit an important early stage in our journey where we collectively decided we needed to pause, put down the prompts, and reflect on AI’s relationship with people, planet, and production, before we pushed blindly ahead.

And sustainability was also just one of the concerns that made us stop and reflect. In an industry that has become reliant on fast-paced, automated systems, is defined by overproduction and overconsumption, and is now rushing headlong towards agentic decision-making, we found ourselves questioning what work can be done by AI, and what work should remain human-centric.

That process is still ongoing, because while we both see a tremendous opportunity in creating a digital platform that can reduce waste, prioritise people and planet, and also output a unique and fantastic product at the end, our perspectives also differ pretty markedly on the use cases for, and the impact of, AI.

We don’t think we’re alone in having these kinds of conversations every day, either. After talking to the team at The Interline, and previewing the survey data that forms part of this report, it feels like there is tension industry-wide between the drive to adopt AI (which is supported by a growing body of evidence), and the hesitance that comes from asking the all-important question: what is the planetary, human, and ethical cost of a prompt?

under+beneath is our brand. It’s led by the two of us —Amelia Tombs and Chloe Grant—combining our different skillsets to create a sustainable, bra-focused fashion tech business, founded on a revolutionary data-native approach for design, production, and traceability. Where fragmented, fractured supply chains and production that far surpasses what the market can sustain are common, we’re working to cut textile waste through automation, customisation, and complete transparency into a fully digital value chain.

One of us, Amelia, is the operational founder. Her focus is on return on investment, efficiency, speed, structure, and precision. The other, Chloe, is sustainability and creative-minded, and comes to AI already deeply concerned about the effect of AI on communities and climate.

We are both committed to helping  fashion leave the days of opaque overproduction behind, embracing a data-driven, accessible era, where cutting edge tools solve legacy issues that currently have immense repercussions for people and our environment alike. 

Where we differ is our perspective on how far AI should be embedded in that commitment. So, to help other brands wrestling with the same conflicting ideas, we each want to set out what we believe.

Amelia: From an operational perspective, as the more tech-driven of the duo,I have found AI to be instrumental to my logistical approach to running under+beneath. 

From automating admin tasks, to speeding up crucial pre-seed investor outreach, AI is my helpful companion.  It removes admin debt, which typically slows down start-ups in the crucial early days, and allows me to take on the work of multiple people. 

I use AI on a day-to-day basis, from drafting emails to helping code the brand’s platform. Access to AI has allowed me to take minimal coding skills and transform them into an MVP-ready environment, backed up by code and architecture co-written with Gen AI. Alongside speeding up the establishment of under+beneath, and allowing milestones to be reached quicker and more efficiently, this intervention of AI has also been a major cost-saver for an early start-up. 

Whilst AI can never replace the experience of an industry professional, learning alongside coding agents has provided me with a front-row seat to the company’s tech development in a way that I would not have had if I’d taken a typical non-technical founder approach and had an engineer create work that was, at least in some way, opaque to me. In this way AI can take on the form of a learning tool—an invitation to incorporate new skills into our brand’s repertoire.

As a neurodivergent founder I have also discovered AI to be incredibly useful as a tool to accelerate the business. By engaging with AI—learning how its ‘brain’ works, uncovering how to communicate with it, and decoding its outputs—I have, in turn, gained a deeper understanding of how my own brain works, and what steps can be put in place to create a more accessible workplace for me and others. These accessibility  measures could include introducing an AI assistant into deep research projects, or using it to draft concise, meaningful emails that eliminate the need for people to interpret weighted symbols and complex emotions. 

By staying deeply aware of how my brain processes things, I believe I have found the true advantage with AI: not using it for everything, but using it precisely where my processing style or the business encounters friction.

This does not mean that I’m planning to simply stay on the bleeding edge of AI, and to progressively incorporate more capabilities into my own personal workflow, or into the brand, without interrogating why I’m doing it. My aim is to consciously evaluate our internal use of AI by always asking: does it give us back  more meaningful time? Does its use compromise what makes us distinct? Will it impact our individuality or creativity as a start-up? These are the things I need to weigh up against the fact that AI has become a fundamental part of how I work as an individual.

I also feel that being mindful and considered is necessary, even for people who have embraced AI and obtained value from working with it. There is a tendency today for people to ‘automate first, ask questions later,’ and I firmly believe this distracts from AI’s game-changing capabilities, as well as accelerating and perpetuating parts of the business model that should be challenged rather than automated. 

If we fall into the trap of instilling AI as the default, and accepting its potentially problematic outcomes, then I believe we’re missing an opportunity to meaningfully decrease overproduction, cut waste, and improve operating margins and efficiency. We shouldn’t just point AI at an existing model and ask it to go faster; we should be using it as an opportunity for real industry transformation, in sustainability and beyond.

Chloe: As the design-focused founder, I take a different stance to Amelia on the benefits of using AI every day. 

I class myself as someone who is openly anti-Generative AI, because I question its immense environmental impact, and, as a creative and writer, I worry about the influence it will have on the future of design and free-thinking.

I’m aware these are bigger, more philosophical issues than the near-term benefits my Co-Founder, and plenty of other people, are observing, but my fear is that we focus so much on day-to-day productivity that we don’t notice ourselves losing something essential.

For that reason, we don’t use AI for creative purposes at under+beneath.

I want to reiterate, though, that AI can still be useful in the production process. Applying automated, data-driven principles to a custom-fit design flow allows for the creation of unique patterns, tailored to the individual, produced through a series of parametric, machine-learning algorithms. This is not AI-powered creativity, but AI-powered logistical power. 

Building game-changing tools such as this—that actively solve an industry problem and strive to reduce vast amounts of textile waste through conscious, on-demand production—is where AI can truly shine, because nothing is lost and there’s everything to potentially gain.

So my philosophy is this: rather than relying on Gen AI as a dynamic design crutch, or as a cheap shortcut for generating buyer-facing images, we should be questioning where AI will actively help address fashion’s crisis points. How can it sit alongside authentic creativity, rather than detract from it? Where would automated processes shine without negatively impacting design integrity, the people making our clothes, or the environment? 

The other vital consideration is that AI has a heavy counterweight on the other side of adoption. The environmental impact of AI is well-documented, with AI-related water usage predicted to hit between 4.2-6.6 billion cubic metres by 2027*. That’s equivalent to half the annual water usage of the entirety of the United Kingdom, and it’s especially troubling when we consider that water scarcity, planet-wide, is already at a crisis point. According to research by the UN, we do not need to look as far forward as next year to see the looming impact: in 2025, data centres consumed 4.5 trillion litres of water, which is “enough to meet the needs of more than 600 million people in Sub-Saharan Africa, while [also] generating 189 million tons of carbon dioxide emissions.”

My driving-force for co-founding under+beneath is to decrease the impact of the fashion industry on our planet, not contribute further to its environmental decline. With this in mind I encourage the adoption of AI sparingly, and only in parts of the industry that could truly benefit from machine-driven inputs – i.e the parts that can, themselves, generate a direct sustainability return that net offsets the impact of using AI in the first place. Suitable use-cases could include: automated Digital Product Passports (that provide direct, transparent access to information), AI-powered textile sorting facilities that actively tackle the textile waste crisis, digital identification software for unlabelled textile waste, and automated pattern creation for localised, on-demand production.

I don’t think I’m alone in believing that AI should work in harmony with the vision of a circular fashion future, making sustainable options more accessible and speedy, rather than generating a whole host of new issues for an industry already deep in crisis. In a world inundated with clothing and fashion-related content, brand owners must ask the question: do we truly need AI to generate even more of it? Or, instead, can we use it to find, secure, and then scale a more sustainable model?

Reflections

Combining these two opposing views and usage of AI—with one of us openly embracing it, and the other becoming increasingly more wary—has helped turn under+beneath into a dynamic start-up. One that is not distracted by the newest shiny thing, but is instead building on a foundation of research, authenticity, and desire to change the fashion landscape. Innately questioning our ethical AI standpoint and defying the fashion tech status quo to create a start-up that truly considers people and planet at every stage.

Balancing both of our views on AI has resulted in a company that explores the possibilities of an AI-driven tech world, led by Amelia, but is then grounded by the thoughtful, planet-focused mindset of Chloe, questioning: ‘how does this align with our values?’. 

The combination of these two standpoints has resulted in a start-up that is embracing a data-native approach, without spiralling into a business that is wholly reliant on AI and devoid of ethical creativity. 

Whilst simultaneously pushing sustainable inspiration to the limit—creating fashion tech that actively responds to the industry’s problems—without feeling the need to be penned in by the current parameters of the circular fashion model or fashion’s digital-phobic production. Reframing our difference of opinion as a competitive advantage: Amelia pushes the business forward, Chloe holds it accountable.

No, we are not an AI-native brand, despite getting our start in the AI era. But being AI-native, unqualified, feels like as much of a risk as shunning the technology completely.

In practice, we want to turn this pragmatic perspective into complete transparency. Our vow is that no automated features will be used that trade our brand’s integrity for marginal efficiency, but where profound efficiencies can be found and scaled, helping to challenge an entrenched business model, we see those as valid. 

We are committing to maintaining authentic storytelling and images that are free from Gen AI or re-touching, because it’s important for a brand like ours to be relatable.

We describe ourselves as a human-centric brand. We want to prioritise people at each and every stage of the design and production process, and that means more than just keeping a human-in-the-loop. We are combining custom-fit technology with a transparent, people-focused supply chain, to uplift a fragmented industry, and implement AI in tactical, useful ways. Ensuring that the people who wear our bras, the people who make them, and the planet, remain front and centre.

The point of AI, as we see it, is not to generate content and ideas around the margins of a broken business model, but instead to generating a new era for bras: powered by data-native fashion tech and grounded in sustainability. You don’t have to ignore AI to do this, but you don’t have to blindly embrace it either.

---

*Li, P., Yang, J., Islam, M. & Ren, S. (2025). ‘Making AI Less “Thirsty”: Uncovering and Addressing the Secret Water Footprint of AI Models’, Communications of the ACM, 68(7), pp. 54-61. https://doi.org/10.1145/3724499