New Look adopts Fermat AI for product design

Retail adopters of AI visualisation reduce cycles of physical sampling and move product-development effort earlier into digital asset creation, which raises the importance of accurate rendering, colorimetry, metadata and PLM integration for practitioners. These changes shift engineering work toward pipelines that ensure versioning, consistency, and traceability between AI-generated visuals and downstream manufacturing specifications.

Multiple outlets report that New Look announced a partnership with AI visualisation platform Fermat to equip its buying and design teams with AI-powered tools to create virtual product renders, test design iterations and explore different prints, colourways and styling options before physical samples are produced (Drapers; FashionNetwork; Retail Gazette; RetailTechInnovationHub). The retailers' coverage quotes Creative Director Anica Wislawski saying Fermat lets designers "bring ideas to life faster, explore more creative possibilities and refine products through multiple iterations before they reach the customer" (FashionNetwork; Retail Gazette; RetailTechInnovationHub). Retail Gazette additionally reports that insights from Club New Look are being used across the business and that Club New Look has passed 1,000,000 members.

For practitioners, the core technical implications are predictable and concrete. High-fidelity AI visualisation workflows require reliable material and texture models, consistent color profiles, and calibrated lighting to keep virtual samples useful for fit, trim and manufacturing decisions. Image-generation alone is not sufficient; teams typically need integration layers that attach product metadata (size, fabric composition, construction notes) to visual outputs so renders map to bill-of-materials and PLM records. Asset management, canonical identifiers, and schema for variant handling become gating concerns when multiple digital iterations are produced rapidly.

Industry-pattern observations: Retail implementations often surface three operational gaps:

• •the need for governance over generated assets to prevent proliferation of unapproved variants
• •QA processes to validate that virtual colourways match printed or dyed fabrics in production
• •instrumentation for designer feedback loops so the model improvements reflect real user preferences rather than transient stylistic artifacts. These are generic patterns observed in other fashion AI rollouts and do not assert New Look's internal priorities

Editorial analysis: Practitioners should note common issues vendors and adopters encounter: intellectual-property entanglements when models are trained on third-party images, potential mismatch between rendered and manufactured textures, and the requirement to log provenance for regulatory or brand-compliance reviews. Where loyalty and customer-data signals (here, Club New Look) inform design choices, data governance and bias mitigation processes matter for ensuring that automated suggestions do not overfit to noisy segments.

Monitor three indicators after rollout: whether Fermat outputs are integrated into New Look's PLM or PIM systems (coverage that cites integration will indicate maturity), how the company measures virtual-to-physical fidelity (published metrics or trials), and whether the partnership includes tooling for version control and asset provenance. Public reporting of those items would come from vendor case studies or follow-up pieces in trade press.

The New Look-Fermat announcement, as reported across Drapers, FashionNetwork, Retail Gazette and RetailTechInnovationHub, is an example of a broader pattern where mid-market retailers adopt AI visualisation to accelerate product cycles and reduce sampling waste. The practical implications for data scientists and ML engineers lie less in the generative models themselves and more in the engineering work needed to make generated assets deterministic, auditable and manufacturable.