Sustainability teams are spread thin. Everything from reporting and compliance to developing responsible sourcing strategies and chasing data from hundreds, if not thousands, of sources crosses their desks. Even at multi-million-dollar brands, these teams are often just a handful of people, forced to spend more time with their heads buried in spreadsheets than pushing the kind of agenda-setting sustainability strategies they got into this work to implement. It’s no surprise that the promise of hyper-efficient AI seems so seductive.
Industry insiders say AI can ease the strain on sustainability teams and their supply chain partners alike by automating environmental reporting, improving data quality, repackaging it for different sources and verifying traceability. There are also potential efficiency gains in the supply chain itself through smarter material use and more accurate demand planning. “It’s certainly having a lot of positive outcomes around helping sustainability teams [carry out] reporting, allowing them to focus much more on programming than compliance,” says Annie Agle, vice president of impact and sustainability at US outdoor brand Cotopaxi.
But there’s a catch. The true impact of organizational AI use remains largely undefineds. Without careful trackings and measurement, brands could unwittingly increase their footprint via the very technology they are relying on to help decrease it.
Generative AI could help fashion companies streamline their supply chains and save emissions, but the technology is energy intensive. Experts warn against jumping on the bandwagon.
“The upsides are very apparent, but we also know there are negative impacts that are not understood yet,” says Agle. “We don’t know the implications of that digital footprint on our GHG [greenhouse gas] measurement.”
To understand how sustainability teams are embedding AI into their operations and how they are approaching such unknowns, Replica Handbag Store Business spoke to a small sample of brands representing different sizes and market segments. While some are years into AI usage, others are still in the experimental stage, but it’s clear that all consider AI has a role to play in their business going forward.
How Sustainability Teams Are Using AI
H&M says it uses AI across supply chain, logistics, marketing, sales and customer experience, and that AI is supporting its commitment to only produce what it sells. It says it does this by increasingly harnessing the power of AI to optimize how many products to make, where to sell them and when. “It has positive effects on resource consumption, but also in terms of inventory, raw materials and emissions,” the brand explains.
Luxury group Kering, which owns brands including Gucci and Balenciaga, appointed Pierre Houlès as chief digital, AI and IT officer in March 2026. It has been deploying AI across select houses for several years. Like H&M, it is using analytical AI to help forecast demand and optimize inventory levels for each product, says chief sustainability and institutional affairs officer Marie-Claire Daveu. And like Cotopaxi, it is leveraging AI to automate and improve the reliability of reporting, using tools that automatically collects data from Kering sites and intelligently correct it. At the product development level, its Italy-based Material Innovation Lab, which researches and incubates more sustainable materials, developed an eco-design AI agent to provide technical guidance for its design teams. This helps to bridge often siloed design and sustainability teams to ensure the output of one doesn’t compromise the goals of another.
While H&M and Kering have built AI into key parts of their operations over several years, for other brands, AI is mostly about lightening the load. “When it comes to AI, we’re still in the early stages of exploration,” said Everlane CEO Alfred Chang in a statement shared with Replica Handbag Store Business. “We’re focused on how it might support day-to-day internal processes and time-consuming administrative tasks.”
Similarly, Agle says she and her small team are using AI to translate raw supplier data into GHG measurements, to repackage data in different formats for the many reports they must produce, and to help create visualizations of company impact for wider team comprehension. “It really does help with productivity. I definitely think companies that don’t adopt AI are going to struggle to compete commercially,” says Agle.
Maximilien Abadie, deputy CEO of French technology company Lectra, agrees. Lectra offers a range of AI solutions that can help support fashion brands in meeting their sustainability goals, such as verifying traceability data and optimizing fabric cutting to reduce over-ordering. But Abadie says the golden opportunity is shortening the time to market. “What is at stake for any fashion company in the world? It’s how to be competitive in a world which is shaken on a daily basis, in a world with so many uncertainties, in a world where you cannot predict and forecast what would happen in three months, six months from now, but you need to be present at the right time with the right product, at the right price, with the right quantities in front of the right consumer.”
The reality is that while AI is being leveraged to cut waste and optimize supply chains, it is also being used extensively to fulfil commercial and creative needs. “Since 2018, we have developed more than 15 internal machine learning platforms, all built to enhance creativity, efficiency and customer service,” says Jordi Alex, chief information technology officer at Spanish retailer Mango.
Mango developed most of its platforms in-house, according to Alex. This includes its AI-powered assistant Iris, used to respond to customer queries, handling more than 7.5 million enquiries per year, according to figures shared by the brand. Likewise, its customer-facing tool Gaudi draws upon user browsing and buying habits to generate personalized product recommendations, and Mango also uses generative AI to create campaign and collectsion imagery. Such extensive use cases beyond sustainability increase the urgency for brands to drill down on the underlying impacts.
Unknown Impacts
It is widely acknowledged that the growth of AI has created increased demand for energy and water (used for cooling data centers). A 2025 paper published in Nature stated that at current growth rates, AI servers across the US alone would generate 24-44 million metric tons of carbon dioxide (CO2), equivalent to adding 5-10 million cars to US roads. Their annual water footprint ranges from 731 to 1,125 million cubic meters, compounded by the fact that many data centers have been or are being built in the world’s driest areas. In April 2025, the International Energy Agency (IEA) reported that data centers (which house the IT infrastructure needed to train and deliver AI services) accounted for 1.5% of global energy consumption in 2024, set to rise to 3% by 2030.
While data centers don’t solely power AI, it is estimated that AI workloads will represent half of all data center capacity by 2030, demonstrating its significant, and growing, energy demand. US AI company Anthropic has even acknowledged that the grid capacity and data center growth its technology requires will increase electricity prices for consumers and has stated it will cover infrastructure costs. The company did not respond to requests for comment.
While broader projections can be made, it’s much trickier to calculate the impact of AI use on an organizational level. “It’s going to be very complicated to understand the impacts of AI as a non-AI company. When it’s your service and you own the data centers, you can count the electricity and water usage needed for cooling,” says Cotopaxi’s Agle. But when you’re a fashion brand and you don’t know all the ways in which every single employee is using AI — whether they’re writing emails with ChatGPT or gathering data with internal AI tools — arriving at an exact figure is extremely difficult. It’s also reliant on AI companies measuring and validating their impacts so third parties can use the data for their own calculations. But major players such as OpenAI, Perplexity and Anthropic have not publicly disclosed any emissions data. None of the companies responded to requests for comment.
Daveu says Kering closely monitors the environmental impact of AI across its IT activities, including data centers and cloud usage, and that IT represents less than 2% of its total carbon footprint. Among the companies contacted by Replica Handbag Store Business, Kering was the only one to share an outline of how it is mitigating the impacts of AI use: by prioritizing the use of simple, resource-efficient models and working with technology partners to support decarbonization and increase carbon reporting transparency.
Kering didn’t specify where it sources the data for its calculations, and currently, there are no standardized metrics. To ensure companies can make comparable assessments of the impacts of AI in the future, the Coalition for Sustainable AI — initiated by the French government, the UN Environment Programme (UNEP) and the International Telecommunications Union (ITU) — is calling for global standardization. To that end, in February 2026, the ITU, the UN agency for digital technologies, released guidelines for assessing the environmental impact of AI systems and minimizing and mitigating impact. And in January 2026, the Taskforce on Nature-related Financial Disclosures (TNFD) released draft sector guidance for the technology and communications sector, which aims to help organizations carry out assessments of their nature-related dependencies, impacts, risks and opportunities. These are the first steps toward a more standardized system, similar to those that currently allow brands to release comparable data around the impact of production or transportation, for example.
Until robust frameworks are established and widely adopted, brands will have to forge their own way forward. Agle says that, in lieu of global standardization for calculating AI’s footprint, Cotopaxi will be undertaking “back of napkin math” to try and understand its impact. “Step one is understanding our usage. What contracts do we have between a company and an AI provider? What kind of usage are we seeing from employees? If we say ‘we have X number of employees using Claude, for instance, and Claude is reporting this set of emissions’, what percentage of those emissions do we own? It’s going to be really challenging to measure it, but we need to try,” she says.
From Measuring to Mitigating
For Cotopaxi, measuring impact is step one of creating a responsible AI policy. Some brands already have them in place. H&M developed a responsible AI framework in 2018, while the adoption of AI at Mango is led by a steering committee. “Through centralized governance, training and programs such as AI Champions — internal ambassadors who support adoption and share best practice across teams — we ensure AI is adopted in the right way and supports our people to realize their potential,” says Mango’s Alex. Though many existing policies and frameworks acknowledge the potential negative impacts of AI, they tend to target the ethics of AI, such as safety, anti-bias and transparency, rather than environmental factors.
The French Ministry of Ecological Transition’s general framework for frugal AI use is designed to fill that gap, offering guidance on best practices and how to move from measurement to mitigation. A source from the ministry explains that some of the considerations brands can make include asking whether AI is really necessary for certain tasks, training AI during times of the day data centers are using renewable energy, using tailored, high-quality databases for training to reduce the computing demand and exploring whether a less energy-intensive AI model could be used.
Though sprawling, multitasking AI platforms that rely on data centers have become the dominant picture of AI, there are different options. UK AI company Literal Labs, a spinout of Newcastle University, is betting on creating efficient AI models and tools that don’t need expensive, specialised Graphics Processing Units (GPUs) for processing, but can instead run on smaller, cheaper and more efficient chips and processors that you find in items such as television remotes or microwaves. It can achieve this by ditching the complex mathematics behind algorithms like ChatGPT (known as a neural network) and replacing it with “if/then” statements (known as a logic-based network), which require less computing power.
The company’s chief product officer, Daniel Dykes says the company has not, to date, discovered any trade-offs as both neural networks and logic-based networks can carry out a process called deep learning, allowing them to process complex data. When testing demand forecasting for a food company, the logic-based network was 7% more accurate than a well-known, neural network-based competitor.
“If you can say an algorithm doesn’t require hardware, it doesn’t require a data center, it doesn’t require a new power plant, you’ve solved a lot of problems that present-day AI is posing,” says Dykes. The company, which is targeting EMEA and working in sectors including water utility, claims it is over 50 times more energy efficient when training a model than a neural network counterpart, and 54 times faster.
UK-based DeepGate is also innovating AI that can run on smaller chips, testing it to efficiently detect “wake words” (phrases used to activate tech such as “Hey Siri” or “OK Google”) and classifying images. While still highly capable, it does not represent a wholesale alternative to more demanding systems. Broad-ranging, multi-tasking platforms such as Claude or ChatGPT need to run on GPUs, but many specific platforms can be run on cheaper, more energy-efficient hardware and infrastructure such as DeepGate’s.
Such tasks might include visually assessing fiber or material quality or identifying anomalies in certification documents or supply chain datasets; targeted, specialized workloads. Those simpler systems, explains DeepGate co-founder Luke Taylor, may coexist with bigger systems which would carry out more complex responses or analysis — stepping in only when needed rather than running continuously, which is extremely inefficient.
There are solutions on the horizon, but in order to apply them strategically, brands first need to get to grips with where AI’s impacts lie to ensure its use represents a net gain for sustainability. “It doesn’t make sense to say I improve something on one side, but I’m polluting more and CO2 emissions are increasing drastically [on the other],” says Lectra’s Abadie. He believes that more understanding will lead to shorter iterations and, ultimately, improved AI. “The more we move forward, the more I think people will realize that the impact must be factored in from the beginning. If it creates more negative impact, then it’s not worth investing in it.”
