MSP Art of Business for Sustainable Fashion 2025 Conference

Reducing Waste in the Fashion Industry through Generative AI Powered Consumer Behavior Simulation


Full Abstract (Page 10)

The Problem

Every year, up to 40% of clothing remains unsold, creating massive environmental waste. Our solution uses Generative AI + Vision Transformers to simulate consumer behavior before a single piece of clothing is produced. By predicting what people will actually buy, brands can reduce waste, cut costs, and plan smarter.

Current Industry Methods

Historical Data Forecasting

Provides only basic insights into past performance, cannot capture the exact features and design elements that consumers prefer.

Sample Production

Creating physical samples contributes to material waste and extends production timelines, making the process inefficient

Fast Fashion

Fast fashion is the least sustainable way of fashion planning, involving companies rapidly creating small batches of clothing to meet changing consumer demand.

Methodology

Our aim was to explore whether a GABM+ViT system could simulate consumer reasoning before production, making fashion planning more sustainable and streamlined.

Our Technological Edge

*

Our Technological Edge *

Generative Agent Based Modeling

We created dynamic virtual consumer profiles that reflect consumers’ real-world behaviors and preferences

Vision Transformer Descriptions

Our system automatically breaks down clothing into standardized descriptors—necklines, sleeves, colors, textures, seasonality. This allows direct comparison of design features across thousands of items.

Scalable & Sustainable Testing

Instead of producing physical samples, we run thousands of digital trials instantly. This eliminates material waste and accelerates design decisions, making the entire fashion cycle more efficient.

Implications

  • Fashion Planning

    Test new designs digitally before production

    Pinpoint which design elements resonate with your target customers

    Cut down lead times while increasing design precision

  • Sustainability

    Reduce the environmental impact of fabric, dyes, and energy wasted on unwanted clothing.

    Scale sustainable practices using synthetic data agents to test designs without harm

    Promotes a shift toward a low waste fashion cycle

  • Retail and Business

    Gain predictive insights instead of relying only on past sales

    Reduce costs tied to failed product launches and unsold stock

    Enter new markets more confidently with survey-driven customer segment models