Fifteen dimensions of agent-readiness across the sector, read once a week.
The fashion sector has reached a telling moment: the interaction layer and brand perception are running ahead of the operational depth beneath them. LVMH's VivaTech 2026 showcase is the clearest case. Louis Vuitton's ChatGPT-integrated app and Dior's Generative Engine Optimization workstream are real, deliberate, and consistent with heritage identity. They are also interface and discovery moves. They place a conversational Decision Surface in front of the client without yet rebuilding the decision loops behind it. That gap defines the sector this week.
The binding constraint is scaling. The State of Fashion 2026 figure, now attributed directly to McKinsey and Business of Fashion, is stark: up to 90% of transformative AI projects remain stuck at pilot. Content, discovery, and sizing produce documented wins because data is abundant and errors are cheap. Etro's 46% e-commerce revenue growth from an AI content pipeline is the model case. The Compiled Corporation test asks a harder question: how much of the firm's core decision-making is now automated. On that measure most houses have automated a workstream, not a decision loop. Scaling maturity moves down two points this week for exactly this reason.
Two forces are doing the real work underneath. The first is agent-readiness. Machine-readable product data is becoming the visibility layer, with 45% of commerce leaders ranking it a priority investment and agentic shopping adoption projected to more than double to 46% by year end. A catalogue an agent cannot parse is a sale that goes elsewhere. Fashion's attribute depth, size normalization, and absence labelling make this harder here than in most verticals, and the houses that solve it first will own agent-referred demand.
The second is governance arriving on a deadline. The Chambers Fashion Law 2026 guide documents brands building frameworks for training-data provenance and bias, and the August 2026 AI Act labelling rules plus the EU destruction ban turn AI forecasting and imagery into dated obligations. This is where the Identity Control Surface becomes concrete. As agents touch clienteling, imagery, and forecasting, houses need governance over non-human identity and data lineage, not as ethics decoration but as compliance.
The cohort picture is uneven by design. Kering and LVMH disclose structured programs: Kering's Talent Match platform and Copilot rollout lift workforce and talent readings on genuine evidence. Richemont sits mid-signal. Chanel and Hermès are privately held and scored on thin public evidence, a limitation the reader should hold in mind: their silence is not an absence of activity, it is an absence of disclosure.
The diagnosis for the houses reading this: the interface investments are landing, and perception is earned. The work that remains is unglamorous. Clean, machine-legible product data. Decision loops rebuilt rather than augmented. Agent identity governed before agents transact. The sector that closes the gap between its front end and its back end will convert quiet-revolution credibility into compounding value.
The pilot trap is now directly attributed: up to 90% of transformative AI projects in fashion remain stuck before production. Single-house production wins like Etro do not lift a sector where stalled pilots dominate. The two-point move is held for editorial review.
Machine-readable product data moved to a priority ranking for 45% of commerce leaders, and Dior's Generative Engine Optimization workstream shows a pinnacle house acting on agent-legible discovery. Intent still outpaces deployment, so the move is a measured one point.
Product-AI outcomes sharpened this week: virtual try-on lifting conversion 40% and cutting return costs up to 50%, and AI-referred traffic converting 60% higher. Value is real but gated by data quality.
Where the sector stands on the fifteen-dimension framework. Reference, not the headline. The movement above is the story.
The Agentic Fashion Report scores the fashion vertical on the Applied AI Index 15-dimension enterprise AI-readiness framework (0-100 across Organization, Brand, and Product, five dimensions each, equally weighted within category). Scores derive from public signals: analyst and trade coverage, house communications and filings, and machine-readable commerce evidence. A sector cut of the Applied AI Index methodology.
Produced by Applied Identities using specialized AI analysis. All scores based on publicly available data. Internal deployment data not captured. This analysis is independent and does not represent the views of any company named herein.