What does The State of Fashion 2026 say about AI?

Its theme 'The AI Shopper' reports that 23% of consumers primarily use generative AI to discover products, that ChatGPT drove 16% of Zara's inbound referral traffic in mid-2025, and that agentic commerce could be worth $3 trillion to $5 trillion by 2030. It also says up to 90% of transformative AI projects are stuck at pilot stage.

Where does the $150 billion to $275 billion fashion AI figure come from?

From McKinsey's March 2023 article 'Generative AI: Unlocking the future of fashion', which estimated generative AI could add that much to apparel, fashion and luxury operating profits within three to five years. It is a model, and it is not from a State of Fashion report.

How many consumers let AI agents buy for them?

Very few. BCG's October 2026 study found 81% of US consumers expect to use AI agents when shopping but only 4% would let one complete a purchase alone, and agent referrals were under 1% of site visits in late 2025.

Are companies getting value from AI?

A minority. BCG found 5% of companies capture value at scale and 60% get hardly any material value; McKinsey's 2026 survey found 37% attribute any EBIT impact to AI; Bain reports 10% to 25% EBITDA gains, but only among tech-forward leaders.

Do consultancies think fashion will be bought by agents?

Less than other categories. BCG says fashion, beauty and luxury are less exposed because discovery plays a larger part, and State of Fashion 2026 says a proprietary agent matters less for luxury brands than for retailers.

Fashion AI by the numbers: what consultancies claim

The consultancies agree on direction and disagree on size. McKinsey and The Business of Fashion's State of Fashion 2026 names "The AI Shopper" as a defining theme, and McKinsey models agentic commerce at $3 trillion to $5 trillion globally by 2030. BCG, Bain, Deloitte, Accenture, Capgemini and Forrester all say discovery is moving into AI first. But the same firms' surveys show how far behind the projections reality is: BCG found only 4% of consumers would let an agent complete a purchase alone, Capgemini found 25% had actually used gen AI shopping tools in 2025, and up to 90% of transformative AI projects are stuck at the pilot stage according to State of Fashion itself. For fashion and luxury, BCG and McKinsey both argue the category is a special case, where discovery, emotion and the brand's own advisers matter more than delegation.

This is part five of our fashion AI research series and the fourth of five deep dives. It sets out what each firm claims, with the report and page where we could find one, and separates projections from surveys and surveys from measured results.

What does The State of Fashion say about AI?

The BoF-McKinsey State of Fashion reports are the most cited source on fashion AI, so it is worth being exact about what they say.

The State of Fashion 2025 (November 2024) made "Discovery Reinvented" one of its ten themes. Its executive survey ranked product discovery and customer search as the top generative AI use case, chosen by 50% of fashion executives, ahead of marketing (45%), design and creative (41%) and curated recommendations (39%); store operations came near the bottom at 17% (p. 42). It told executives to make sure product content is optimised for AI search (p. 40).

The State of Fashion 2026 (November 2025) went further with "The AI Shopper". Its key figures (pp. 39 to 46 of the report):

23% of consumers primarily use generative AI to discover products, while traditional search still drives about 80% of search traffic (p. 40).
Shopping-related generative AI traffic grew 4,700% from July 2024 to July 2025, an Adobe figure (p. 40).
ChatGPT accounted for 16% of Zara's inbound referral traffic and 8% of H&M's between June and August 2025, per SimilarWeb (p. 40).
Third-party content such as reviews, blogs and affiliate posts makes up about 80% of the sources AI uses (p. 43).
The case for a proprietary shopping agent is "generally stronger for retailers than brands", and for luxury a proprietary agent "may be relatively less important" (p. 43).
Agentic commerce could be worth $3 trillion to $5 trillion by 2030 under moderate adoption, and brands risk losing ownership of the customer relationship (p. 44).

The same report carries its own counterweight: in "Efficiency Unlocked" it notes that up to 90% of transformative AI projects are stuck at the pilot stage (p. 83), and in "Workforce Rewired" that 92% of organisations will increase generative AI investment while only 1% call their rollouts mature (pp. 31 to 32). For jewellery brands, its "Jewellery Sparkles" theme projects jewellery unit growth of 4.1% a year from 2025 to 2028, four times the rate of clothing, and tells executives to emphasise personalisation (p. 51).

A note on provenance. The widely quoted estimate that generative AI could add $150 billion to $275 billion to apparel, fashion and luxury operating profits within three to five years comes from McKinsey's March 2023 article "Generative AI: Unlocking the future of fashion", not from a State of Fashion report. It was a model written four months after ChatGPT launched. The State of Fashion 2027 had not been published when we wrote this.

What else does McKinsey claim?

"LLM to ROI" (August 2024) modelled $240 billion to $390 billion of value for retailers, a 1.2 to 1.9 point margin uplift. Its survey of 52 Fortune 500 retail executives found 82% had piloted customer-service gen AI but only 36% were scaling, and just two reported full implementation across the organisation. It also estimated a chatbot needs a 2% to 4% basket uplift to justify its costs, a useful threshold for any pilot.
"The agentic commerce opportunity" (October 2025) sized US B2C agentic commerce at $900 billion to $1 trillion by 2030, and named vertical agents such as "fashion stylists" among the models.
"When AI meets desire" (May 2026) argues the luxury prize is "interpretive authority", not maximum automation: the brand, not a general agent, should decide how it is described. Its luxury-shopper figures rest on 31 respondents, which the authors call directional; among them, only 9% wanted fully autonomous agents. It cites Ralph Lauren's Ask Ralph and Zegna's adviser copilot as examples.
"The state of AI in 2026" (August 2026, n=1,719) found nearly nine in ten organisations use AI, about two in ten are scaling agents, and 37% attribute any EBIT impact to AI.

What does BCG say?

BCG has published more on fashion and retail AI than any other firm in our research, and its numbers are the most useful for calibrating hype.

Value is concentrated. In 2024, BCG found only 26% of companies had moved beyond proof of concept to generate value, and 4% created substantial value. In 2025 it found 5% "future-built" and 60% reaping hardly any material value.
People are most of the work. "The AI-First Retailer" (November 2025) puts about 70% of the effort of an AI transformation in people and change management. Its AI at Work 2025 survey of more than 10,600 employees found frontline regular use had "stalled at 51%".
Fashion specifically. "The AI-First Fashion Company" (December 2025) describes agents across design, fit simulation, assortment and pricing, citing Walmart's fashion division cutting production time by up to 18 weeks and Tiffany & Co. using AI-generated client summaries for clienteling. It gives no dollar or margin figure.
Consumers use AI, but rarely let it buy. BCG's survey of more than 13,000 consumers (August 2026) found about 31% use AI at least occasionally when shopping, triple the level of 18 months earlier, and that AI introduced shoppers to a brand they would not otherwise have considered in about 63% of AI-assisted journeys. Its October 2026 agentic commerce study found 81% of US consumers expect to use agents, but only 4% would let one complete a purchase alone; agent referrals were under 1% of site visits in late 2025.
Fashion is "less exposed". The same study says fashion, beauty and luxury are less exposed to agent-led shopping because discovery plays a larger part, and that 91% of retailers fear losing the customer relationship.
Many do not measure. BCG's 2026 CPG and retail survey found more than half of executives do not measure the ROI of consumer AI.

What do Bain, Deloitte, Accenture and Capgemini say?

Bain. The Bain-Altagamma luxury study (January 2025) found personal luxury goods fell to €363 billion in 2024 and the market lost about 50 million customers in two years, and urged brands to use technology and generative AI as accelerators, without AI numbers. Bain's Technology Report 2025 reported 10% to 25% EBITDA gains for tech-forward leaders from scaled AI, while most companies remain "stuck in experimentation mode" and errors compound across multistep agent tasks.

Deloitte. Its 2026 Retail Industry Outlook, a survey of 330 retail executives, found 81% believe generative AI will weaken brand loyalty by 2027, and that AI referrals are 15% to 20% of referral traffic for some retailers. Its State of AI in the Enterprise (n=3,500) found 28% report revenue gains from AI: "Speed is becoming common. Deeper transformation is not." In 2024 Deloitte predicted 25% of enterprises using gen AI would deploy agents in 2025, rising to 50% by 2027; its 2026 predictions cite Gartner's forecast that over 40% of agentic AI projects could be cancelled by 2027.

Accenture. Its "Empowered consumer" research (April 2024) found 74% of consumers had abandoned a purchase because they felt overwhelmed by choice, and 78% wanted AI advisers to remember their preferences. Its Technology Vision 2025 found 80% of executives say similar-sounding chatbots make differentiation hard, and 95% say a consistent AI "personality" will matter for customer-facing agents.

Capgemini. Its 2025 consumer survey (n=12,000) found 58% prefer generative AI tools to search engines for product recommendations. Its 2026 edition found 25% actually used generative AI shopping tools in 2025, 76% want clear rules for when an AI assistant acts, and 71% worry about how their data is used. Its agentic AI research found only 2% of organisations had deployed agents at scale, and trust in fully autonomous agents falling from 43% to 27% in a year.

Forrester. Its predictions for 2027 say one in 20 US and UK consumers will replace traditional shopping with conversations, and one in five users will bypass brand websites as AI does the research.

How should you read the consultancies' numbers?

Three kinds of number appear in these reports, and they should never be added together.

Kind	Examples	How to use it
Projection	$150bn to $275bn (McKinsey 2023); $3T to $5T agentic commerce by 2030 (McKinsey 2025); $240bn to $390bn for retail (McKinsey 2024)	Direction and scale of interest, not a forecast of your results
Survey of intent or attitude	81% expect to use agents (BCG); 58% prefer gen AI for recommendations (Capgemini); 56% intend to shop with gen AI on Black Friday (BCG)	Openness, which runs well ahead of behaviour
Survey of behaviour or outcomes	25% used gen AI shopping tools (Capgemini); 31% use AI when shopping (BCG); 4% would let an agent buy alone (BCG); 37% see any EBIT impact (McKinsey)	The closest to what is happening now

Two more cautions. Several headline figures are other people's data repeated: the 4,700% traffic growth is Adobe's, cited by both BCG and State of Fashion. And sample sizes vary from thousands to a few dozen; McKinsey's luxury-shopper figures rest on 31 people and its 2024 retail-executive survey on 52.

The sharpest disagreement is about scale. McKinsey's $3 trillion to $5 trillion and BCG's expectation that more than half of e-commerce spend could be agent-assisted by 2030 sit against BCG's own 4% for autonomous purchase and Forrester's one in 20 for 2027. "Agent-assisted" and "agent-completed" are different measures, and headlines blur them. The ROI debate splits the same way: leaders' self-reports (Bain's 10% to 25% EBITDA gains) pull up, the share of pilots reaching the P&L (BCG's 5%, State of Fashion's 90% stuck in pilot) pulls down.

Where do the consultancies agree?

Discovery first, delegation later. AI is changing how people find products now; handing over the purchase is rare.
AI visibility is the new search optimisation. Rich, structured product data and third-party content decide whether AI recommends you.
The customer relationship is at stake. State of Fashion, BCG and Deloitte all warn that brands could lose ownership of the customer to AI intermediaries.
Fashion and luxury differ. Discovery, emotion and advisers matter more; the case for autonomous agents is weaker.
Trust limits adoption. Shoppers want clear rules and control over their data.
People and measurement decide ROI. Change management is most of the work, and many companies do not measure.

What should a fashion or jewellery brand take from the consultancies?

Win discovery with your data. Make product data rich and machine-readable, and earn the third-party content AI reads. Part eight covers the fields.
Own the interpretation of your brand. McKinsey's luxury work argues the brand should decide how it is described and sold. A brand-governed stylist or adviser copilot does that; a general-purpose agent does not.
Design for assist, assemble and authorise. Few shoppers want an agent to buy alone. Keep people in the loop and make the rules visible.
In jewellery, lead with personalisation. State of Fashion expects jewellery to outgrow clothing and names personalisation as the priority.
Measure from day one. McKinsey's own threshold, a 2% to 4% basket uplift to justify a chatbot, is a fair bar for any stylist pilot. Part nine shows how to measure it.

Ateliq's design reflects several of these points: its stylist speaks in the brand's own name and voice and answers from the brand's own stock, its agents draft while people approve, and shoppers are remembered only if they opt in and confirm. The consultancies' central caution applies to it as much as to anything: a pilot should be judged on measured results against a control, not on a market projection.

Next in the series: what Google, OpenAI, Amazon, Shopify and the model labs are building.

Fashion AI by the numbers: what McKinsey, BCG, Bain, Deloitte and the analysts claim