Fashion is spending more on AI. The evidence of returns is still thin

In two years, artificial intelligence has gone from a side topic to a board priority in luxury. In 2024, just 5% of luxury houses and groups surveyed by the Comité Colbert and Bain & Company listed AI among their top three priorities for the next three years. In the 2026 edition, the share was 22%. Yet the same report found that nearly 60% of luxury companies have not yet seen a significant effect from any AI deployment.

That gap between ambition and evidence runs through the whole industry. Spending on AI is rising, by every survey measure available, but almost no fashion company tells investors how much it spends on AI specifically. Where returns are reported, they are mostly operational: faster content, fewer returns, better forecasts. Revenue and profit effects attributed to AI remain largely unaudited, and the most optimistic numbers tend to come from technology vendors' own surveys.

This matters now because 2026 budgets are being locked in as consumer behaviour shifts. The BoF and McKinsey State of Fashion 2026 report notes that ChatGPT accounted for 16% of Zara's inbound referral traffic between June and August 2025, according to SimilarWeb data. Companies that cannot measure what their AI spending returns will struggle to decide where the next euro should go.

How much are fashion companies spending on AI?

The honest answer is that nobody outside the companies knows, because the large groups do not disclose it. AI spending sits inside wider investment programmes. Inditex guided ordinary investments of around EUR 2.3 billion for 2026, focused on stores, new technologies and online platforms, according to FashionUnited. H&M plans capex of SEK 9 to 10 billion for its 2026 financial year, mainly for its store portfolio and technology infrastructure, Modaes reported. Kering said at its capital markets day in April 2026 that it would reinvest 5% to 6% of revenue to support organic growth, while describing a technology hub built on cloud systems, agentic AI and digital twins. None of these figures isolates AI.

The groups do describe where the money goes. H&M chief executive Daniel Ervér said the company wants to use AI to forecast demand much more accurately, down to each stock node, and is giving design teams AI tools that he said “can really enhance their creativity on a large scale”. Inditex highlighted Zara's AI try-on tool, which by 31 January 2026 was active in 43 markets and had generated more than 7 million sessions. Both are usage and intent disclosures; neither company attached a cost or a financial return to them.

The most specific disclosure we found came from Poland's LPP, owner of Reserved and Sinsay. At an investor day in May 2026, the company said it spends about 10% of its IT budget on new technology, including AI, Reuters reported. That is a share rather than an amount, and it bundles AI with other new technology.

Surveys fill the gap, with caveats. In BCG's IT Buyer Pulse Check, the net share of fashion and retail respondents planning to increase IT spending over the next 12 months (those increasing minus those decreasing) rose from 26% in the first half of 2023 to 51% in the second half of 2024, moving the sector from sixth to second of nine industries, behind only finance. NVIDIA's third annual State of AI in Retail and CPG survey, published in January 2026, found that 90% of respondents planned to increase AI budgets in 2026, and half expected increases of 10% or more. NVIDIA sells AI computing infrastructure, and the survey covers retail and consumer goods broadly, not fashion alone.

Consultancies add estimates of what the money could buy, drawn from client work rather than audited results. In a November 2025 briefing for fashion and luxury executives, BCG said brands with AI-enabled experiences or their own shopping agents see conversion lifts of 5% to 15%, that using agents in marketing can cut marketing production costs by 30% to 50% and speed up content development by 40%, and that AI in design can reduce sampling costs by 20% to 30%. These are ranges from BCG's own analysis and company examples, including Zalando, not survey averages for the sector, and the document does not report what share of fashion companies achieves them.

Is AI now a top priority for fashion executives?

Yes, though it competes with tariffs, weak demand and pricing. In the BoF and McKinsey State of Fashion 2026 executive survey, 13% of respondents named AI, digital and technology capability development and its integration into operations as the single biggest opportunity for 2026, the most common answer. A year earlier, expanding AI use cases ranked second, at 9%. More than 35% of executives said they already use generative AI in areas such as online customer service, image creation, copywriting and product discovery.

Luxury has moved faster than its reputation suggests. Besides the jump from 5% to 22% in houses ranking AI a top-three priority, the Comité Colbert and Bain report found that 39% of groups and maisons have a defined AI vision, strategy and roadmap, and a further 48% have a vision with some pilot projects. The executive survey behind those figures is small, 35 respondents from 23 maisons and groups, so each percentage point represents a handful of companies.

How many fashion companies have scaled AI beyond pilots?

Fewer than the priority rankings imply. Speaking about the State of Fashion 2026 research in February 2026, McKinsey's Colleen Baum said the industry is “a bit slower to adopt AI” and that “about 90 percent of projects have stalled at the pilot phase”. The report itself cites wider McKinsey research across industries in which 92% of organisations say they will increase investment in generative AI, while only 1% describe their rollouts as mature.

Where luxury houses have scaled AI, it is mostly in functions customers never see. The Comité Colbert and Bain survey found scaled adoption in knowledge management at 74% of respondents, in IT at 41% and in sourcing, procurement and manufacturing at 33%.

Customer-facing AI is less advanced. Only 5% of luxury houses reported large-scale digital personalisation with measurable impact, and 5% reported significant impact from scaled AI-enabled CRM. In AI-augmented sales assistance, the picture is more mixed: 35% of maisons have deployed solutions with at least some impact, 20% are developing or piloting, and 45% have no plans or are still evaluating.

What returns do fashion companies report from AI?

The independent survey evidence is sober. Nearly 60% of luxury companies in the Comité Colbert and Bain study have yet to realise significant impact from any AI deployment, and even in the most advanced functions only 13% to 15% report significant business impact. Respondents named the obstacles: 60% cited difficulty defining a value-oriented strategy and 60% a lack of internal resources, Modaes reported.

Cross-industry data point the same way. McKinsey's State of AI 2026 survey of 1,719 participants, fielded in May and June 2026, found that 44% of respondents say AI is scaling across their enterprise, but only 37% say it has contributed positively to EBIT, essentially unchanged from 2025. So-called high performers, which attribute at least 5% of EBIT to AI, make up about 6% of respondents. Cost reductions are most often reported in supply chain management, service operations and manufacturing, revenue gains in marketing and sales.

Size makes a difference. In the same McKinsey survey, 54% of respondents at organisations with at least USD 1 billion in revenue reported enterprise-wide scaling of AI, against one-third at smaller organisations, and 40% of large organisations are scaling AI agents in at least one function, up from 27% a year earlier. Respondents in consumer goods and retail most often report using agents in marketing and sales. For mid-sized fashion brands, that suggests the gap with the largest groups is widening rather than closing.

Vendor research is far more positive. In the NVIDIA retail and consumer goods survey, 89% of respondents said AI is helping to increase annual revenue and 95% said it is helping to reduce annual costs; 30% reported revenue increases above 10% and 37% reported cost reductions above 10%. The survey reports “hundreds of responses” without describing its methodology, and the figures are self-assessed. They are best read as a signal of sentiment among companies already invested in AI, not as measured returns.

Which fashion companies can show measurable AI returns?

Two European retailers stand out because they attach numbers to specific processes. Zalando's full-year 2025 release of March 2026 said the company scaled AI-generated product content from near zero to 90% in one year, cut campaign creation time from six weeks to days and lifted content output by 70%. Its Size and Fit AI, using measurements from more than one million customers, cut size-related returns by more than 8%, and its matchmaking models drove a 13% increase in items added to bags. The State of Fashion 2026 report adds that generative AI cut Zalando's image production costs by 90% and production time from six to eight weeks to three to four days.

LPP gave the most direct cost figures. According to Reuters' report from its May 2026 investor day, AI now generates 80% of the group's marketing visuals, up from 20% in 2025, and content production costs have fallen by 60%. Trend analysis and collection design take 6 to 12 weeks, down from 6 to 12 months. LPP also uses an AI tool that analyses foot traffic and mobile data to choose store sites, and says 98% of physical Sinsay stores are profitable, something it described as impossible without the location analysis.

Both cases share features. The returns are concentrated in content production, where costs are visible and work is repetitive; in returns and sizing, where the outcome is easy to count; and in planning cycles. Both are company statements rather than audited results, and neither company publishes the cost of the AI that produced the savings, so a true return on investment cannot be calculated from outside.

Why do surveys and company reports tell different stories?

Because they measure different things, from different people, for different purposes. The table sets out the main sources used in this article and what each can and cannot tell a fashion executive.

Self-reported surveys capture belief and intention, which are useful for gauging competitive pressure. Company disclosures capture outcomes, but only those management chooses to publish. Audited financial statements, the strongest evidence, contain no AI line at all. The result is that the industry's most confident numbers are the least verifiable.

What are the risks of spending on AI without clear returns?

The first risk is pilot purgatory. If McKinsey's estimate that about 90% of fashion AI projects stall at the pilot stage is close to the mark, much of today's spending is buying experiments, not capabilities. The Bain respondents' complaints about strategy and internal resources suggest the bottleneck is organisational rather than technical.

The second is workforce disruption without a plan. State of Fashion 2026 cites McKinsey estimates that 28% to 38% of current activities of consumer goods and retail workers in Europe and the US could be affected by technology by 2030, and that 60% of leaders across industries see company culture as the biggest obstacle to tech-related change. Savings in content production, the area with the clearest returns, come directly from creative and production work.

The third is measurement drift. Operational metrics such as content volume can rise while brand quality or conversion falls, and none of the disclosures reviewed here report negative results. Self-assessed revenue effects, such as those in vendor surveys, cannot be separated from price increases, marketing spend or market recovery.

The counter-argument is that waiting is also costly. AI is changing how customers find products. State of Fashion 2026 reports that shopping-related generative AI searches grew 4,700% between July 2024 and July 2025, that 23% of consumers already primarily use generative AI to discover products, and that AI-driven revenue per visit on US retail sites grew 84% between January and July 2025. It estimates that agentic commerce could be worth USD 3 to 5 trillion by 2030 under moderate adoption. Infrastructure spending in data and planning systems may not show returns for years, and companies that wait for proof may find the channel has moved without them.

What should fashion companies do now?

1. Ring-fence an AI budget line Track AI spending separately inside technology capex and operating costs, even if it is not published, so returns can be calculated.
2. Start where costs are visible Prioritise content production, sizing and returns, and demand planning, where the cases of Zalando and LPP show outcomes can be counted.
3. Set a kill date for pilots Give each pilot a defined metric and deadline, then scale or stop it, to avoid the pilot stage where most fashion projects stall.
4. Measure against a baseline Record cost per asset, forecast error or return rates before deployment, so improvements are not inferred after the fact.
5. Treat vendor surveys as sentiment Use independent and audited evidence for investment decisions, and read self-reported revenue gains as indicators, not proof.
6. Plan the workforce shift Decide in advance how content, customer service and planning roles will change, and budget for retraining alongside the technology.