EDITOR'S NOTE
Smooth Operators is dead, long live LUXE ÆTERNAI!
Dear reader, after a few weeks of silence, I decided to "pivot" my newsletter Smooth Operators to focus exclusively on luxury, which has always been my true passion. Agentic AI is the lens, not the subject. Here, I don't chronicle model releases with a footnote saying "and for luxury this means..."
What I'm offering is to start from the houses themselves -- their craft, what they do, what works, what doesn't -- and I analyze it through the eyes of someone who knows the luxury sector and digital inside out. The analytical framework I'm proposing? Luxury rests on five fundamental invariants, the 5 Es: Excellence (the artisan's gesture, absolute quality), Exception (rarity, uniqueness), Experience ("wow moments," sensory immersion), Emotion (brand attachment, storytelling), Elevation (aspiration, personal transformation). By definition, these invariants don't change. What does change is the pressure that agentic AI exerts on each of them. When agentic AI dilutes an invariant, it fails. When it elevates one, it becomes Soulful Tech, as Kapferer puts it. That's the dial I measure here, every Thursday.
KEY MOMENT
On January 11, LVMH took the stage at the NRF (National Retail Federation) Retail's Big Show in New York -- the world's largest retail congress, 40,000 professionals -- to publicly lay out its AI doctrine. The central announcement: "AI for All," an initiative covering all 75 of the group's houses.
This is no longer just another digital project. It's the formalization of a systematic program, presented as such before the entire industry.
Source: NRF.com, January 11, 2026
PARADOX OF THE WEEK
LVMH names its initiative "AI for All" and deploys agents across 75 houses. But its doctrine states that AI must remain invisible so that humans remain central. Present everywhere, apparent nowhere. And meanwhile, the houses that best protected their mystique for decades are discovering that others have been writing about them -- and that those writings now feed the agents that recommend them to their customers.
DEEP DIVE
What NRF 2026 teaches us about agentic AI in luxury
The doctrine is starting to crystallize. And it's more subtle than we thought.
Dimension: Hyper-personalization x Seamless experience -- Customer experience | Brand sovereignty
What agents know about you -- and what you never decided
When Soumia Hadjali, Senior Vice President of Global Client Development and Digital at Louis Vuitton, takes the mic at the NRF Retail's Big Show in January in front of 40,000 professionals, she doesn't present a tool. She defines what AI agents must embody for a luxury house: "anticipate, understand, interpret, respond to the client's intent, and build trust, legitimacy -- all powered by AI." In twenty years in this industry, I would have given the exact same definition of an excellent boutique director -- except that one doesn't take vacations and whose memory never fails.
This is not a technical description. It's a brand position. And it contrasts sharply with how e-commerce platforms talk about agents -- efficiency, conversion, automation. For Louis Vuitton, agentic commerce is "a form of digital concierge, anchored in continuity and emotional coherence."
But behind this doctrine lies a subject that no one in the industry has named yet. What LLMs (Large Language Models) know about luxury houses, no one in those houses decided. These models are trained on third-party corpora -- press articles, resale platforms, collector reviews. And it's this subject -- not the LVMH doctrine itself -- that should concern every digital director in the sector.
"AI for All": what it really means
The initiative covers all 75 LVMH houses. Gonzague de Pirey, the group's Omnichannel and Data Director, is explicit: each house has its own AI transformation plan, with common points identified across commerce, marketing, and operations. The logic relies on MaIA -- LVMH's internal AI platform, built with Google Cloud since 2021, which centralizes house data and powers agents specific to each use case. The infrastructure is shared, but each house's identity remains sovereign.
What's new in the NRF announcement: the shift from "projects by house" to "group-wide initiative." LVMH stops treating AI as isolated experimentation. This is the signal that the pilot phase is officially over at the world's leading luxury group.
De Pirey puts it this way: "Luxury is fundamentally about people -- artistic creation, handcrafting, and the bonds that our advisors create with our clients." AI enhances these connections. It doesn't replace them.
The widening gap -- and a dimension no one has named yet
The State of Fashion 2026 (McKinsey / BoF, Nov.-Dec. 2025) establishes it unequivocally: 90% of "transformative" AI projects in luxury remain stuck at the pilot stage. The barriers are structural -- governance, fragmented data, silos -- as much as cultural.
Harvard Business Review goes further in its February 2026 analysis: when autonomous agents do the shopping instead of the consumer, the balance of power shifts from retailers toward brands and the agents themselves. In a world of shopping agents, what matters is no longer media visibility -- it's machine readability.
What NRF 2026 reveals: the divide is no longer between luxury groups and tech platforms. It's within the sector itself. LVMH formalizes a doctrine, Ralph Lauren deployed "Ask Ralph" in September 2025. Meanwhile, the majority of houses are still debating their data governance.
Having accompanied houses through their digital transformation for twenty years, I can gauge just how telling this figure is: the problem is almost never technical. It's a problem of governance and organizational will. The tools exist. What's missing is the decision to use them -- and the structure to do it at scale.
And time is running out. The RealReal alone represents more than ten years of structured data on the perceived value of luxury pieces: $520 million in sales volume in Q3 2025 alone. A house whose products are excellent but that communicates little can be underrepresented in agent responses -- not because it's less good, but because others have written less about it. In a sector where scarcity produces desire, algorithmic underrepresentation becomes a real commercial risk.
What it means for others
LVMH's lead is not technological. It's organizational. The group has been building its data infrastructure since 2021 -- that's what enables deploying agents across 75 houses today. This is, in my view, LVMH's true competitive advantage in this race -- not the technology, but a five-year head start on data governance. Houses starting this work now are behind on infrastructure, not on tools.
The good news: the tools are accessible. The bad news: without clean data and governance in place, an agent produces generic responses. In a luxury context, a generic response is worse than no response -- it erodes the image.
Since August 2025, Article 53 of the EU AI Act grants rights holders a formal right over the use of their content in LLM training corpora. For houses whose campaigns, editorial texts, and product descriptions constitute a valuable documentary heritage, this right is actionable. To date, none has publicly disclosed any initiative in this regard.
The most advanced operational response came from elsewhere -- from Ralph Lauren, which chose to build its own agent with its own data rather than negotiate its presence in corpora it doesn't control. We analyze this in detail in the case study below.
What really changes: agentic AI in luxury doesn't deploy on a sprint. It's built on infrastructure. And that infrastructure needs to be laid down now -- before others decide for you what agents know about you.
WHAT'S MOVING
The RealReal: record GMV at $520M, the luxury resale market crosses a threshold
November 10, 2025 | The RealReal -- official Q3 2025 results
The RealReal's GMV (Gross Merchandise Value) reaches $520M in Q3 2025, an all-time record, up 20% year-over-year. Revenue grows 17% to $173.6M. Adjusted EBITDA comes in at $9.3M -- positive for the second consecutive quarter. A trajectory that's settling in: two quarters in the black, and an operational base that no longer looks like a gamble.
Why it matters: The luxury resale market is no longer a thesis. It's structurally profitable. Houses without a strategy for this segment are letting others define the perceived value of their pieces -- and their clients consult The RealReal before buying firsthand.
HBR: when agents do the shopping, the balance of power shifts
February 2026 | Harvard Business Review
HBR documents a structural shift: when autonomous agents do the shopping instead of the consumer, power shifts in favor of brands whose content is readable by LLMs. Brands with a coherent, documented, and structured identity gain visibility. Those that relied on ad spend lose their advantage. Bain & Company specifies that 50% of consumers remain cautious about fully autonomous purchases -- the agent advises more than it decides.
Why it matters: This is the argument that should trigger the GEO (Generative Engine Optimization) conversation in every luxury boardroom. It's no longer a question of search ranking -- it's a question of survival in a customer journey that starts with a prompt, not a Google search.
Agentic commerce: $3-5 trillion by 2030, and already an open front
November-December 2025 | State of Fashion 2026 -- McKinsey / BoF
McKinsey estimates the potential of agentic commerce at $3 to $5 trillion by 2030. Shopping searches on generative AI platforms surged 4,700% between July 2024 and July 2025. 53% of American consumers who used generative AI for research also used it to purchase (Q2 2025).
Why it matters: These numbers currently concern the mass market. But they outline the behavior of the next generation of premium buyers. What normalizes at the bottom of the market becomes an expectation at the top. Estimated timeline: 3 to 5 years.
EU AI Act: houses have a right over their training data -- and aren't using it
In force since August 2025 | Regulation (EU) 2024/1689 -- EU AI Act, Art. 53
Since August 2025, Article 53 requires publishers of large language models to maintain a copyright compliance policy on their training data, and grants rights holders an opt-out right over the use of their content. For luxury houses whose campaign imagery, editorial texts, and collection descriptions constitute a valuable corpus, this right is actionable now.
Why it matters: No house has, to my knowledge, publicly disclosed a formal initiative. The Comite Colbert has not taken a collective stance. The regulatory window is open -- and appears still unexploited. To be continued...
CASE STUDY
Ralph Lauren / Ask Ralph: what the first truly shoppable luxury agent teaches about brand sovereignty
Sources: Ralph Lauren official press release, Sept. 9, 2025 | Digital Commerce 360, Sept. 12, 2025
On September 9, 2025, Ralph Lauren launches "Ask Ralph" -- a conversational assistant integrated into the brand's US mobile app, developed with Microsoft Azure OpenAI. You can ask it in natural language to put together an outfit for a specific occasion. Recommendations are directly shoppable from the interface. No detour through the site's search engine. No results list. A conversation, a selection, a purchase. Simple, effective.
This is not a customer service chatbot. It's a commercial agent with a deliberately restricted scope -- and that's exactly where the lesson lies.
The sequence of decisions that matters
Three structural decisions are apparent in this deployment, in this order:
1. Closed scope. The agent only answers questions for which Ralph Lauren has a controlled response: its own catalog, its own style codes. No competitor comparisons, no open-ended questions about the brand. This choice isn't a technical limitation -- it's a governance decision. The agent can't say something wrong about Ralph Lauren because it only has access to what Ralph Lauren has provided.
2. Proprietary environment. Deployment happens within the brand's own app, not on a third-party platform. All conversation data -- what the customer is looking for, what they ask, what they buy -- belongs to Ralph Lauren. No intermediary learns in the brand's place what its customers want.
3. Integrated transaction from Day 1. Most "discovery first" deployments separate recommendation from purchase -- which creates a leak between intent and conversion. Ralph Lauren makes the opposite choice: the agent recommends and closes within the same flow. It's a decision about the value the agent must produce -- and about the fact that this value must be measurable immediately.
What it reveals about brand sovereignty
Ask Ralph is the exact operational answer to the problem raised in the DEEP DIVE. Rather than suffer what LLMs say about Ralph Lauren based on corpora the brand doesn't control, the brand creates its own agent with its own data, in its own environment. It takes back control of the first point of contact between a customer and a product recommendation.
The difference with a generalist LLM queried about Ralph Lauren is radical: one aggregates what others have said about the brand, the other says only what the brand has decided to say.
Key learnings
-- A restricted-scope, well-documented agent is more useful -- and less risky -- than an omniscient, poorly controlled one. Value comes from precision, not breadth.
-- The question "in what environment does the agent live?" is as strategic as "what does the agent do?" Deploying on a third-party platform means giving away customer relationship data to an intermediary.
-- Making the agent shoppable from the start forces you to have clean, up-to-date, and structured product data. It's a constraint that forces the house to solve data governance problems it could have ignored for much longer.
The compass -- 3 questions to ask before launching
Can your agent get something wrong about your brand? If so, the scope isn't defined enough. Narrow it until the answer is no.
Who owns the conversation data? If the answer isn't "we do," revisit the architecture before deployment. Once data is with a third party, it doesn't come back.
How do you measure success at 90 days? Without a KPI (Key Performance Indicator) defined upfront, the agent remains a pilot -- and joins the 90% stuck at that stage. Ralph Lauren had a simple answer: conversion rate from the agent. You should have yours before writing the first line of prompt.
ON MY RADAR -- Weak signal
GEO: from an emerging signal to a $200M market. In under a year.
GEO (Generative Engine Optimization) has reached critical mass in just a few months: 24 competing platforms, over $200 million raised collectively. Profound is named leader in the AEO (Answer Engine Optimization) category on G2, Winter 2026. Goodie AI has just launched an "Agentic Commerce Optimizer" -- a tool dedicated to brand presence in autonomous shopping agents. Google has published in parallel an open protocol, the UCP (Universal Commerce Protocol), to standardize communication between agents and merchant systems.
What this tells us: the market has decided that GEO is a budget line item, not an experiment. And the agencies mastering this terrain are not the ones that managed SEO (Search Engine Optimization) for the last ten years.
For luxury houses whose websites are built for emotion rather than machine-readable structure, the urgency is real. What isn't documented for LLMs doesn't surface in an agent's recommendations. And what doesn't surface doesn't generate a visit.
The question is no longer "should we care?" but "who in your organization is responsible for your visibility in AI responses?"
Sources: Clutch, Feb. 2026 | Google Blog, Feb. 2026
ON MY READING LIST THIS WEEK...
The Adolescence of Technology -- Dario Amodei -- A 20,000-word essay I recommend to anyone deploying AI agents in a professional context. This is not a CEO doing crisis communication -- it's a structured analysis of the systemic risks AI poses to economies, democracies, and labor markets. Uncomfortable to read. Essential.
Annotated full text in French -- Le Grand Continent -- If you don't have time for the 50 pages of Dario Amodei's essay in English, Le Grand Continent provides an annotated French version. The editorial commentary adds real depth -- it's not a simple translation.
What Happens for Luxury If AI Goes Bust? -- BoF, Nov. 2025 -- The question nobody asks at luxury x AI conferences: what happens for major houses if the bubble bursts? Affluent American clients -- the most resilient luxury segment during the current down-cycle -- are directly exposed to an AI-linked market correction. A rarely covered angle of systemic fragility, from BoF's luxury editor.
GOING FURTHER
On GEO -- market overview and tools:
10 GEO platforms compared -- Platelunch Collective, Feb. 2026 -- The most comprehensive analysis of optimization tools for generative engines: 10 platforms scrutinized, no marketing fluff. Starting point for anyone who wants to go beyond the names.
On NRF (National Retail Federation) 2026 -- the LVMH session in detail:
Official NRF 2026 session -- "Where Craft Meets Intelligence" -- The LVMH session page at the NRF Retail's Big Show: speakers, summary, and program context. For those who want to trace back to the source of the "AI for All" doctrine.
On the luxury market and AI -- global overview:
Finding a New Longevity for Luxury -- Bain & Company x Altagamma, Nov. 2025 -- The 24th edition of the benchmark annual study on the global luxury market. What directly concerns this newsletter: house-operated online boutiques are gaining ground thanks to AI-augmented personalization, while multi-brand platforms are retreating. The signal is clear -- customer data sovereignty is starting to pay off.
YOUR TAKE?
That's a wrap for this first edition of LUXE ÆTERNAI. I hope you enjoyed reading it as much as I enjoyed putting it together. This new format is still in its early days, and your feedback is invaluable.
LUXE ÆTERNAI | Every Thursday morning in your inbox.