How Luxury Brands Are Using AI — and Why Most Are Getting It Backwards

TL;DR — Key takeaways

• LVMH’s internal AI agent MaIA handles 1.5 million queries per month from 40,000 employees — proving AI at scale is operational, not creative.

• By mid-2026, AI agents account for 15–25% of luxury travel bookings. The buyer is no longer always human.

• 90% of luxury AI pilots failed to scale — not because of authenticity concerns, but because of poor data foundations.

• The brands winning aren’t optimising AI for human customers. They’re restructuring their brand signals so AI agents can read, rank, and recommend them.

• Epinium’s Agentic Buyer Stack framework maps the three layers luxury brands must build to survive the AI-mediated commerce shift.

Five years ago, the question was whether luxury brands should use AI at all. The fear was real: would automation erode the mystique, the craftsmanship, the deliberate scarcity that makes a Hermès bag worth ten times a comparable leather good? That debate is settled. Every major maison is deploying AI. The new question — and almost nobody is asking it — is whether they’re deploying it in the right direction.

The LVMH Proof Point Most Analysts Are Reading Wrong

When LVMH disclosed that its proprietary AI agent MaIA fields 1.5 million queries per month from 40,000 employees across 24 brands, the coverage framed it as a creativity story. Headlines celebrated AI-generated mood boards, trend synthesis, and client stylists getting real-time inventory intelligence. All true. All beside the point.

What made LVMH’s approach defensible — and replicable — is the underlying architecture: a unified data estate built before the AI layer was switched on. Franck Le Moal, LVMH’s Chief Information Officer, was explicit about this sequence. Data governance first. AI tooling second. Most of LVMH’s industry peers did it in reverse. That is why, according to a 2025 Gartner benchmark, 90% of enterprise AI initiatives in retail failed to scale beyond the pilot phase, with poor data quality cited as the primary cause.

The authenticity argument that dominated 2023 and 2024 coverage was always a distraction. Burberry and Jacquemus use AI for client behaviour modelling and inventory allocation while keeping creative direction entirely human. The tension isn’t AI versus craft. It’s whether you built your data foundation before or after you bought the platform licence. See also our analysis of why most ecommerce AI integrations stall at the data layer — the same failure mode shows up consistently across categories.

90%

of enterprise AI retail pilots fail to scale beyond proof-of-concept

Source: Gartner, 2025

Where the Luxury AI Investment Is Actually Going — and Why It Feels Miscalibrated

The dominant investment pattern is hyper-personalisation for human shoppers. AI stylist tools. Recommendation engines. Dynamic pricing. Virtual try-on. These are real applications with real ROI, and I’m not dismissing them. But when you look at where the structural shift in luxury commerce is happening, the allocation feels off.

In May 2026, Skift reported that AI agents already account for 15–25% of luxury travel bookings globally, with projections toward 40% by 2027. These aren’t bookings where a human used AI to research options and then clicked confirm. These are autonomous agent transactions — an AI, acting on behalf of a user’s stated preferences and budget parameters, selecting and booking the experience without human review of the specific decision.

When the buyer is an AI agent, your brand’s curated Instagram aesthetic is irrelevant. The agent doesn’t see it. What it reads is structured data: product titles, attribute consistency, review sentiment patterns, pricing stability signals, and whether your brand appears in the training sets of the models doing the ranking. A brand that has spent three years refining its visual identity on social media and zero time on its product data architecture is, from an agentic commerce perspective, invisible.

Working with a high-end jewellery brand on their Amazon channel, we saw this exact failure mode play out. Three years of detailed brand guidelines and visual system documentation. Zero time spent on product attribute standardisation. Their AI content tool generated polished descriptions that ranked nowhere, because the underlying feed contradicted itself on material specifications and SKU naming at scale. What we see at Epinium is that brands with wholesale or marketplace channel exposure are disproportionately exposed to this structural risk.

The Agentic Buyer Stack: A Framework for What Comes Next

Here is the Agentic Buyer Stack — a framework we use internally to help brands assess readiness for AI-mediated commerce. Three layers:

Layer 1 — Signal coherence. Does your brand identity resolve consistently across product data, website, social, and third-party retail channels? An AI agent cross-references these signals. Inconsistency reads as low-confidence, and the agent routes to a competitor with cleaner data.

Layer 2 — Attribute completeness. AI purchase agents are trained on structured specifications. A luxury watch brand with incomplete technical attributes — movement type, water resistance, case material — in its product feed is algorithmically disadvantaged against a mid-range brand that filled every field. Completeness beats heritage in the agent’s evaluation function.

Layer 3 — Sentiment stability. Agents weight review consistency over time. A spike in negative reviews — even one that was addressed — creates a confidence penalty in agentic ranking models. Brand reputation management is no longer just a communications function; it has become a commerce function.

According to Retailnews.ai 2026 data, 41% of US shoppers already use AI tools to discover products. The question isn’t whether this accelerates. It’s whether your brand is legible to the systems doing the discovering.

15–25%

of luxury travel bookings already made by AI agents, mid-2026

Source: Skift, May 2026

Richemont’s Right Bets — and Kering’s Structural Weakness

Richemont’s AI roadmap centres on operational excellence and digital provenance — using AI-enhanced digital IDs to fight counterfeiting and manage secondary market pricing signals. Smart, unglamorous work that directly defends margin. Conversational agents on their YNAP platforms surface relevant recommendations while keeping human advisors as the relationship layer. The architecture is coherent.

Kering’s approach is less unified. Its individual maisons — Gucci, Saint Laurent, Bottega Veneta — run largely autonomous AI initiatives without a shared data infrastructure. In practice, Gucci’s AI client data doesn’t inform Kering’s portfolio-level demand planning. That’s a structural inefficiency that will compound as agentic commerce grows, because agents increasingly look for conglomerate-level signals when evaluating brand authority and pricing consistency.

Luxury Brands AI in 2025–2026: What Actually Changed

Agentic commerce crossed the threshold (Q1 2026)

AI agent bookings moved from theoretical to statistically significant. Skift’s May 2026 data confirmed 15–25% of luxury hospitality bookings are now fully autonomous. Brands that hadn’t restructured their distribution data by Q1 2026 are already behind the first adoption curve.

LVMH deployed MaIA to 24 maisons (2025)

The rollout confirmed that centralised AI infrastructure outperforms brand-by-brand deployment. MaIA’s 40,000-user base and 1.5 million monthly queries established a benchmark for scale that no single brand deploying an isolated AI system can match cost-efficiently.

EU AI Act enforcement began (August 2025)

The Act’s transparency requirements hit luxury’s personalisation engines hard. AI-generated product recommendations shown to EU consumers must now carry disclosure labels in specific cases. Several brands pulled AI stylist features from their European apps rather than navigate the compliance ambiguity. This will resolve as guidance clarifies, but brands without legal-AI integration in their tech stack remain exposed in the interim.

Counterfeit detection became an arms race (2025–2026)

Luxury’s counterfeit problem entered a new phase: AI-generated fake product imagery that passes visual authentication. Richemont, LVMH, and Chanel accelerated investments in provenance verification AI — blockchain-linked digital IDs, NFC authentication at point of sale — as the counterfeiting ecosystem adopted the same generative tools brands were using for marketing.

Epinium data

Across brand and manufacturer clients we’ve worked with on Amazon channel optimisation, those who completed a full attribute audit before deploying AI content tools saw an average 34% improvement in organic ranking within 90 days — versus 8% for brands that applied AI content generation to unaudited product feeds. The foundation matters more than the model.

How AI Deployment Differs Across Luxury Tiers

Brand Tier	Primary AI Use	Data Foundation	Agentic Readiness
Mega-conglomerates (LVMH, Richemont)	Internal ops, demand forecasting, fraud detection	Centralised, governed	High
Independent heritage brands (Hermès, Chanel)	Client relationship intelligence, supply chain	Siloed by function	Medium
Accessible luxury (Michael Kors, Coach)	Marketing automation, trend forecasting	Fragmented	Low–Medium
Luxury manufacturers / wholesale	Content generation, listing optimisation	Often incomplete	Low

FREE SESSION

Is your brand legible to AI buyers?

In 30 minutes, we map your brand’s agentic readiness across product data, distribution channels, and AI search visibility — and tell you exactly where the gaps are.

Get your AI readiness audit → ✓ Free   ✓ 30 min   ✓ No pitch

Frequently Asked Questions: Luxury Brands and AI

Does using AI make a luxury brand less authentic?

The short answer is no — if deployed correctly. LVMH’s own philosophy, “technology everywhere, visible nowhere,” captures the practical rule. AI that surfaces client history for a sales advisor, forecasts demand to prevent stockouts, or detects counterfeit listings is entirely invisible to the end consumer. Authenticity is eroded when AI generates creative output without editorial judgement, not when it optimises operations behind the scenes.

What’s the difference between how LVMH and Kering use AI?

LVMH centralised its data infrastructure first, then deployed AI across all 24 maisons via MaIA. Kering let individual brands run separate AI initiatives without a shared data layer. The practical consequence: LVMH can run portfolio-level demand forecasting and cross-brand client intelligence; Kering cannot at the same resolution. Both are investing heavily, but the architectural decision will compound over the next three to five years.

What is agentic commerce, and why does it matter for luxury?

Agentic commerce refers to AI agents that complete purchases autonomously on behalf of users, without human review of each transaction. For luxury, this matters because AI agents evaluate brands differently than humans do: they parse structured data, consistency of brand signals across channels, and review sentiment patterns. A brand that optimised entirely for human-facing aesthetics and neglected its product data architecture is at a structural disadvantage when AI becomes the buyer.

How should luxury brands respond to the EU AI Act?

The immediate priority is mapping which AI systems touch EU consumers and whether those systems qualify as high-risk under the Act’s definitions. Personalisation engines and recommendation systems are currently in a grey zone pending guidance clarification. In the interim, brands should document the logic of recommendation systems, implement human oversight layers, and consult legal teams with both AI regulation and luxury sector experience. Pulling features entirely — as some did in 2025 — is an overreaction that will cost competitive ground.

Can a small luxury brand with no IT team implement AI meaningfully?

Yes, with two caveats. First, start with a data audit, not an AI tool. Understand what product data you have, where it lives, and how consistent it is across channels. Second, the highest-ROI first step for most smaller luxury brands isn’t a custom AI system — it’s ensuring your product feed is complete and correctly structured for platforms that AI agents and search systems already crawl. The infrastructure work is less glamorous than the AI tool, but it’s what the AI actually reads.

What about AI and counterfeit detection for luxury brands?

This is one of the most concrete ROI cases for AI in luxury right now. Richemont and LVMH are both deploying AI-linked provenance systems — NFC chips, blockchain-anchored digital IDs — that authenticate products at point of sale and track them into the secondary market. The challenge accelerated in 2025 when generative AI allowed counterfeiters to produce convincing fake product imagery that defeated visual authentication. Brands without a digital provenance layer are increasingly exposed on resale platforms.

Is AI-generated content appropriate for luxury brand communications?

For internal documentation, briefing, and operational content: yes, with editorial review. For client-facing creative — campaigns, brand storytelling, hero imagery — the risk is a noticeable drop in emotional resonance that luxury consumers detect even if they can’t articulate why. Research from the Boston Institute of Analytics confirms that luxury consumers react negatively to disclosed AI-generated advertising. The practical rule: use AI to prepare and iterate, keep humans in the final creative decision.

What does Ralph Lauren’s AI stylist failure actually teach us?

The Ask Ralph experiment revealed a fundamental misunderstanding of luxury personalisation’s purpose. Luxury curation means narrowing choice — presenting one perfect option with authority and context. AI stylist tools optimised for recommendation surface area produced technically competent but emotionally empty suggestions. For luxury brands building AI-assisted client tools, the success metric is not recommendation volume. It is recommendation authority: does the AI say “this is the one” with enough context that the client trusts it?

How do AI agents rank luxury brands when making purchase decisions?

Based on current evidence from agentic commerce research, the key ranking factors are: structured data completeness (how fully specified is the product across all attributes), signal consistency (does the brand identity resolve coherently across all touchpoints the agent can access), sentiment stability (what is the review trajectory over time), and price signal clarity (is pricing consistent or erratic across channels). Heritage, visual identity, and brand narrative — the things luxury spends most on — are not directly legible to the agent. That’s the gap most brands haven’t closed.

What’s the single most important AI investment a luxury brand should make right now?

A data foundation audit — not a technology purchase. Map every product across every channel. Identify where attributes are missing, where naming conventions are inconsistent, and where your brand signal diverges between your direct site and your wholesale partners. Once that’s clear, the right AI tools become obvious. Buying a platform first — before completing that audit — is precisely why 90% of pilots fail. The model doesn’t fix the data problem. It amplifies it.

The luxury industry spent three years debating whether AI belonged in the same sentence as craftsmanship. That conversation consumed boardroom time that could have gone into building the data infrastructure that now separates brands with genuine AI leverage from those with expensive experiments. The next three years will be less philosophical. Agentic commerce doesn’t care about brand heritage. It reads attributes, consistency signals, and sentiment trajectory. The maisons that treat that reality as an operational problem to solve — rather than a creative threat to manage — are the ones that will hold the ground.

TRANSFORM BY EPINIUM

Make your brand legible to the AI buying the next generation of luxury.

Brands working with Epinium’s Transform practice have closed the agentic readiness gap in under 90 days — across product data, distribution channels, and AI search positioning.

Book a free session →

Free · 30 min · No commitment