by Sohom Banerjee
annually. The European Commission has also estimated that 4-9 per cent of textile products placed on the EU market are destroyed before ever being used. These are not merely failures of recycling; they reveal a deeper problem. Waste is often built into the economics of the industry.
Artificial intelligence (AI) is frequently presented as part of the solution. That claim deserves both attention and caution. AI can improve demand forecasting by combining sales data with weather, online searches, local preferences, returns, pricing behaviour and emerging trends. In principle, this allows firms to replace large speculative production runs with smaller initial batches and rapid replenishment. AI-enabled forecasting and inventory systems have been associated with potential inventory reductions of 20-30 per cent. The implication is significant. The cleanest garment is often the one that never needed to be produced.
But the practical question is not whether brands deploy AI; it is what they instruct it to optimise. An algorithm rewarded solely for revenue, conversion or speed may identify trends earlier, shorten design cycles and encourage still more consumption. The same technology that reduces inventory risk can therefore accelerate the very model responsible for excess. Efficiency, without restraint, can simply make overconsumption more efficient.
Firms therefore need to redesign both operating models and performance metrics. Demand-forecasting systems should incorporate “do-not-produce” thresholds when confidence is weak. Management should track unsold inventory, markdown intensity, returns and material waste per unit of revenue alongside conventional commercial indicators. AI investment should be judged partly by whether it reduces absolute material throughput and dependence on virgin fibres, not merely whether it improves sell-through rates.
A second weakness is visibility. Fashion supply chains stretch across farms, fibre producers, dye houses, factories, logistics providers and retailers. AI cannot optimise what companies cannot reliably measure. Supplier-level data on fibre composition, water and energy use, defects, waste and end-of-life outcomes therefore need to become interoperable and auditable. The European Union’s movement towards Digital Product Passports for textiles and extended producer responsibility rules offers an important direction. Other jurisdictions should adopt comparable standards without creating compliance systems that inadvertently exclude smaller suppliers.
Circularity requires an equally practical approach. Recycled fibres accounted for only 7.6 per cent of global fibre production in 2024, while less than 1 per cent came from recycled pre- and post-consumer textiles. AI-assisted sorting, hyperspectral imaging and computer vision can improve fibre identification and textile-to-textile recovery. Recommendation and pricing algorithms can also strengthen resale, repair and rental markets. Yet these technologies will achieve little if garments continue to be designed from blended materials that are difficult to separate, repair or recycle. “Design for circularity” must precede “AI for circularity”.
Policymakers consequently have a role beyond merely encouraging technological innovation. They should require credible disclosure of unsold and destroyed stock, establish measurable waste-reduction targets, strengthen producer-financed collection and recycling systems, and develop common standards for product-level environmental data. Regulators should also demand evidence behind AI-enabled sustainability claims; otherwise, algorithmic efficiency risks becoming another vocabulary for greenwashing.
The real opportunity, therefore, is not to make fast fashion technologically sophisticated while leaving its economic logic untouched. It is to use intelligence to impose discipline on an industry historically organised around abundance. AI can help fashion predict better, produce less, trace more, recover value and extend product life, but only when commercial incentives, regulation and technology point in the same direction.
Algorithms will not cure fashion’s addiction to excess on their own. But if they are used to optimise restraint rather than velocity, they may help rewrite the industry’s governing logic. From produce, push and discount to predict, prove and recirculate.
About Author
,Sohom Banerjee founder of Quantive Advisory LLP and the author of the book “Governing Artificial Intelligence”.
Sohom Banerjee – Sohom Banerjee is a public policy and technology strategist, AI governance expert, author, and founder of Quantive Advisory LLP, an advisory and research firm focused on artificial intelligence, digital transformation, and technology policy,
