The AI scorecard: what 15 fashion groups have actually disclosed about AI

LVMH's internal generative AI agent, MaIA, now handles more than 2 million requests a month from 40,000 employees, according to a Wall Street Journal report cited by PYMNTS in June 2025. That single figure is more than most of the world's largest fashion groups have disclosed about their entire use of artificial intelligence.

To see who is doing what, we built a transparent scorecard of 15 large fashion groups across luxury, sportswear, fast fashion, denim and online retail, rating only what they or credible press reports have made public since January 2024. The answer to the headline question is uneven: customer-facing chat and styling tools are now common, AI in planning and supply chain is disclosed far less often, and measured results are rare. Only one group, H&M, has published a detailed AI governance framework.

This matters now because investors, regulators and employees are starting to ask the same question. Models and photographers are contesting synthetic imagery, employees want to know how their work will change, and AI assistants are beginning to shape how consumers find products. A company that cannot show what it does with AI, and under which rules, will struggle to argue that it is ahead.

How did we score fashion companies on AI?

We rated each group from 0 to 3 in five areas: design and content (product ideation, imagery, copy), planning and supply chain (forecasting, assortment, pricing, production), customer-facing AI (assistants, stylists, sizing, search), employee tools (internal assistants and agents) and governance (published principles, policies or controls). The evidence window runs from January 2024 to October 2026; earlier launches are mentioned but not scored.

• 0: no specific public evidence found in the sources we reviewed. This is a statement about disclosure, not proof that a company does nothing.
• 1: a general statement, a data initiative without AI detail, or a basic control.
• 2: a named AI deployment or pilot with a clear purpose.
• 3: a deployment at scale with a disclosed usage or result figure, or for governance a published framework applied to all AI projects.

We used company press releases, earnings call summaries and reports in quality trade and business press. We did not use vendor marketing claims unless confirmed by the company, and we did not reward announcements without detail. The scorecard is a snapshot of transparency as much as of capability, and every rating can move as soon as a company publishes more.

Read across the columns, the pattern is clear. Seven groups score 2 or more for customer-facing AI and six for design and content, the two areas where tools are visible and easy to announce. Only four reach 2 in planning and supply chain, three in employee tools, and one in governance. Total scores range from 10 out of 15 for LVMH to zero for Kering and adidas.

Segments do not explain much. Luxury produces both the highest score (LVMH, 10) and the lowest (Kering, 0), with Prada in between on 2. In fast fashion, H&M Group scores 8 while Inditex scores 3 and Fast Retailing and Shein, two of the most data-driven operators in the industry, score 1 each because they rarely describe their methods as AI. Sportswear is the least transparent segment in our sample: Puma scores 4, Nike 2 and adidas zero. What separates the leaders is less their segment than their willingness to name tools, users and results.

Which fashion company discloses the most about AI?

LVMH leads on breadth. PYMNTS, citing The Wall Street Journal, reported in June 2025 that the group uses predictive AI, generative AI and AI agents across supply chain, pricing, product design and marketing. Brand tools summarise past customer interactions and help sales advisers draft personalised messages; pricing tools adjust prices to currency movements; design teams use AI to build mood boards; e-commerce teams generate product copy. The tools run on a central data platform LVMH built with Google over four years.

The group has also explained how it wants AI used. At VivaTech in May 2024, Axel de Goursac, director of LVMH's AI Factory, said: "We don't want to replace but augment humans with artificial intelligence." Cointelegraph reported that MaIA, launched in early 2024, is based on OpenAI's GPT-4, available only to employees and was developed with Stanford University's Institute for Human-Centered Artificial Intelligence. LVMH trained 1,500 employees in AI in 2023. What is missing is a published AI policy and any figure on savings or sales from these tools.

Franck Le Maol, LVMH's group IT and technology director, framed the ambition in PYMNTS's account: "Tech is, for me, mandatory to become super-efficient and at the same time keep the spirit and the essence of luxury."

Which companies report measurable results from AI?

Very few, and that is the most important finding of the scorecard. Zalando is the clearest example. At its annual results in March 2026, the Berlin-based platform said its AI-based size advice and recommendation tools had raised average items per order by 13% in 2025 and cut returns by 8%, Reuters reported. Zalando stores body measurements for more than one million customers and adds about 20,000 a week. Co-chief executive David Schröder said: "AI is more than a tool; it drives growth and improves efficiency." Zalando shares rose about 13% at one point on the day, their largest daily gain in two years.

Tapestry, owner of Coach and Kate Spade, offers the second case. Its in-house platform Mira, built partly on Amazon Bedrock and covered by Tapestry's first AI patent, pulls data from across the business and surfaces insights in "seconds to minutes", replacing manual work across several dashboards that could sometimes take days, according to Digital Commerce 360 in May 2026. Employees use it to plan assortments, manage inventory and track consumer trends. Chief executive Joanne Crevoiserat called it "a powerful tool that puts business insights into the hands of decision makers across the company." Tapestry's 2025 annual report says it is improving predictive AI, generative AI and machine learning in planning, marketing, personalisation, pricing and product creation, and Kate Spade runs a conversational Gift Assistant on its mobile site.

Neither company claims that AI alone explains its numbers. Tapestry's revenue rose 21% to USD 1.92 billion in the quarter to 28 March 2026, and Zalando's 2025 revenue rose nearly 17% to EUR 12.3 billion, but both businesses have many other drivers. The disclosed metrics are narrow and operational. Still, they are the kind of evidence that most of their peers have not provided.

How are fashion brands using AI with customers and in imagery?

Conversational styling is the most common public use. Ralph Lauren launched Ask Ralph in its US app on 9 September 2025, built with Microsoft on Azure OpenAI; it suggests outfits from Polo Ralph Lauren collections based on current stock, with expansion to other brands and markets planned. Levi Strauss & Co. runs an Outfitting feature in its app in the US, Canada and major European markets. ASOS began trialling Styled for You in November 2025, an AI tool trained on more than 100,000 studio outfits, as it reported a 12% fall in annual sales.

Puma took a different route in September 2024, using Google's Imagen 2 on Vertex AI to generate localised backgrounds for product images, for example showing a trail shoe near Mount Fuji to shoppers in Japan, and using Vertex AI Search for Retail on its sites. Puma said the imagery improved click-through rates and time to market but did not publish a percentage. That gap between claim and number is typical of the customer-facing column.

Imagery is where fashion's AI use is most visible and most contested. H&M said in March 2025 that it would create AI digital twins of 30 models for some marketing content, with the models' permission, and published the first images in July 2025. Chief creative officer Jörgen Andersson said: "We're exploring emerging technologies like generative AI to amplify creativity and reimagine how we showcase fashion." In December 2025 Reuters reported that Zara uses AI to edit images of real models into different outfits. Inditex said it uses AI "only to complement our existing processes" and compensates models "in line with industry best practice".

In product design, the evidence is thinner. Nike's A.I.R. project, shown in Paris during the 2024 Olympics, used generative AI to produce early concept images for prototypes with 13 athletes across four sports, which were then 3D printed. Nike vice president Roger Chen admitted the first outputs looked alike: "We noticed that a lot of the AI images interpreting Air were bound by a similar fluid aesthetic." LVMH's mood-board tools and Tapestry's reference to product creation complete a short list.

Is AI being used in fashion planning and supply chains?

Less visibly than in marketing, and often without the word AI. Prada Group announced in April 2025 that it would use o9's AI-based planning platform to approve leather goods production orders, build finished and semi-finished production plans and align raw material purchases across internal factories and external suppliers. No results have been reported yet. H&M Group describes machine-learning demand forecasting that uses sales history, trends, weather and local events, according to CTO Magazine's account of the company's public communications.

Two of the world's largest apparel businesses talk about data rather than AI. Fast Retailing, owner of Uniqlo, built a Management Cockpit that gathers online product reviews and customer service feedback to improve products and build demand forecasts, Inside Retail reported in September 2025, citing Nikkei; the report does not mention AI. Shein's Peter Pernot Day told the Retail Technology Show in London in April 2025 that the company runs a "unique, demand-based supply chain" fed by marketplace feedback, without attributing it to AI. Both score 1 in planning, reflecting disclosure rather than capability.

Who is giving employees AI agents?

Internal agents are the newest category. Besides LVMH's MaIA, Levi Strauss & Co. announced in November 2025 that it is building a "super-agent" with Microsoft inside Teams, coordinating sub-agents for IT, human resources and operations, with rollout planned from early 2026. Its STITCH assistant already answers store staff questions on products, procedures and training in 60 US stores.

Tapestry's Mira also belongs here. Chief data and analytics officer Fabio Luzzi said "Mira is the result of years of deliberate investment in our data infrastructure", a point that applies across the scorecard: the groups with the strongest disclosures are those that built central data platforms first.

What are the limits of an AI scorecard?

A disclosure scorecard measures what companies choose to say, not what they do. Groups such as adidas and Kering may run extensive AI programmes without publicising them; adidas's third quarter 2025 earnings release contains no reference to AI, which is a fact about the release, not about its operations. Kering's ChatGPT-based shopping assistant Madeline on its KNXT platform dates from 2023 and falls outside our window. Conversely, a well-publicised pilot can score 2 even if it never scales.

• Company claims are not audited. Zalando's 13% and 8% figures and Tapestry's time savings are company statements, not independently verified results.
• Secondary reporting. Several ratings rely on press summaries of calls or reports, which may compress nuance.
• Silence can be strategic. AI-edited imagery has drawn criticism from photographers; the Association of Photographers in London warned that it would reduce commissions. Some brands may prefer not to disclose.
• Snapshot in time. The scorecard covers January 2024 to October 2026 and will change with each annual report season.

Governance is the weakest column. H&M Group's Responsible AI Framework sets out nine principles, from fair and transparent to governed and secure, and requires every AI project to pass a checklist, according to CTO Magazine's account; the company also promised that its synthetic model images would be labelled and watermarked. Inditex's public commitments on models, Tapestry's role-based access controls for Mira and LVMH's employee-only, privacy-preserving MaIA earn a 1. No other group in our sample has published comparable rules.

What should fashion companies do now?

1. Publish an AI use register. List the main AI systems in design, planning, customer service and operations, with their purpose and owner, as H&M has started to do with its principles.
2. Report one hard metric per use case. Follow Zalando's example of items per order and returns, or Tapestry's time to insight, rather than general claims of efficiency.
3. Build the data platform first. LVMH and Tapestry tie their AI tools to central data platforms built over several years; agents without clean data add little.
4. Set rules for synthetic imagery. Agree consent, pay and labelling terms with models and photographers before scaling AI-generated or AI-edited content.
5. Move AI into planning. Pilot AI in forecasting, assortment or production planning, as Prada is doing in leather goods, where margins are decided.
6. Equip staff, not only shoppers. Store and office assistants such as Levi's STITCH or LVMH's MaIA reach thousands of employees and are easier to measure than consumer chatbots.