Glossary Agentic AI × Luxury.

The key concepts for understanding what agentic AI changes for luxury maisons. Drawn from the editions of the LUXE ÆTERNAI newsletter, enriched as audits are delivered.

The Agentic AI × Luxury glossary gathers the essential concepts for understanding how AI agents — programs able to browse, compare, book and buy autonomously — are transforming the relationship between Luxury Houses and their clients. Each term is defined, then illuminated with context specific to Luxury.

A

23 terms

  1. A2A (Agent2Agent)

    An open protocol enabling AI agents built by different vendors to discover one another, delegate tasks and exchange data securely. Launched by Google on 9 April 2025 and handed to the Linux Foundation on 23 June 2025, it relies on standard web building blocks and "agent cards". Seven founders (Google, AWS, Cisco, Microsoft, Salesforce, SAP, ServiceNow); more than 150 member organisations in 2026.

    Luxury contextA2A is the grammar that lets a house's internal agents (client advisory, inventory, logistics) converse with those of its partners (payment, carrier) without bespoke development. The challenge specific to luxury: retaining control over what its agents expose externally, since interoperability must never dilute the brand's control over the experience and customer data.

  2. ACP (Agentic Commerce Protocol)

    An OpenAI × Stripe standard for commerce within AI agents. An open protocol allowing agents to discover, compare and buy products.

    Luxury contextLuxury houses hesitate to join it (loss of control). DLG offers an alternative infrastructure (LuxuryIQ MCP).

  3. AEO (Answer Engine Optimization)

    Optimization for answer engines: structuring content so that a generative engine (AI Overviews, ChatGPT, Perplexity, Copilot) extracts, cites and attributes the brand within its answer, rather than ranking it among a list of links. The goal is no longer the click but inclusion in the synthesized answer. Often presented as the "answer" building block of the broader whole that is GEO.

    Luxury contextFor a luxury house, being the answer rather than a link becomes the new currency of visibility: when a client asks ChatGPT for a rare woody fragrance, AEO determines whether the brand is named. This calls for rewriting product pages and content in natural language and for tending to third-party sources, all the more so since brand sites carry little weight among the sources actually cited by AI engines.

  4. Agent Card

    A JSON document standardised by the A2A protocol that serves as an agent's "digital business card". It declares its identity, its service address, its skills, its authentication schemes and its technical capabilities. Published at a standard address (/.well-known/agent-card.json), it allows other agents to discover it and verify compatibility before any exchange.

    Luxury contextThis is the identity record that makes a house's agent visible and queryable by the ecosystem: before triggering anything, a partner agent reads the card to learn what the agent can do, and under what conditions. For a brand, it is the opportunity to precisely frame what its agent exposes to the outside world and what it keeps private.

  5. Agent identity

    A distinct identity account assigned to an AI agent, separate from those of humans, so that it can authenticate and obtain authorizations on an organization's systems. In the Microsoft Entra model (2026), the agent holds no secret of its own: it is created from a blueprint, tied to a responsible human sponsor, and receives only limited, explicitly consented access tokens.

    Luxury contextWhen an agent acts on its own (placing an order with a supplier, adjusting a price, reading a client calendar), one must know who did what and on whose behalf. Agent identity traces every action, designates the accountable human, and makes it possible to revoke the rights of an entire family of agents at once. A governance prerequisite, sensitive as soon as the agent touches the file of ultra-high-net-worth clients (UHNWI).

  6. Agentic browser

    A web browser embedding an AI agent capable, from a high-level instruction, of understanding a page and chaining actions on its own (clicking, entering data, comparing, adding to cart) instead of waiting for each click. The category took hold in 2025: Comet (Perplexity), ChatGPT Atlas (OpenAI), Edge's Copilot mode, Opera Neon, plus specialised players.

    Luxury contextThis category redefines who visits a brand's site: increasingly, it is an agent, not a human, that reads, compares and adds to cart. The house must keep its site readable and actionable by these agents while preserving the human experience. A corollary issue: these browsers expose users to hidden instruction injections within pages, the dominant security risk for 2026.

  7. Agentic commerce

    A commercial model in which AI agents act on behalf of consumers to search for, compare and buy products.

    Luxury contextLuxury sells emotion and experience. When an agent buys in the client's place, how do you preserve the ceremony of the sale?

  8. Agentic Storefronts

    A Shopify feature (enabled by default in late March 2026) that makes catalogues readable by ChatGPT's AI agents. Stores become interfaces for autonomous shopping agents.

    Luxury contextFor luxury houses on Shopify, the catalogue is now read by AI agents before it is seen by a human. The product page becomes an act of algorithmic communication.

  9. Agentic Token (Mastercard Agent Pay)

    A payment token at the heart of Mastercard Agent Pay, unveiled on 29 April 2025, which extends card tokenisation to agent-driven commerce. It links a tokenised card to a specific agent, a merchant scope and a consent policy (ceiling, authorised merchants, duration). An agent (ChatGPT, Copilot, Gemini) completes a purchase without holding the real card number; the token is revocable agent by agent.

    Luxury contextThe Agentic Token turns the agent into a bounded cardholder: the House knows which agent is paying, within what scope and under what authorisation. On sensitive purchases, confining an agent to a single seller, a single ceiling and a limited duration, with revocation at any time, protects image and margin and reduces exposure to card-not-present fraud and chargebacks.

  10. Agentic visibility

    A brand's presence and actionability in the responses of AI agents. Distinct from traditional web ranking: being cited is not enough. One must be cited accurately, with a link, in the right tone.

    Luxury contextActionability scores vary considerably from one agent to another. A brand can be cited by ChatGPT and ignored by Gemini.

  11. AGNTCY (Internet of Agents)

    An open-source infrastructure project for AI agent interoperability, opened by Cisco in March 2025 with LangChain and Galileo, then hosted by the Linux Foundation on 29 July 2025 (more than 65 contributing companies). It standardises four layers: agent discovery, verifiable identity, messaging (SLIM protocol) and observability, to build an "Internet of Agents" interoperable with MCP and A2A.

    Luxury contextWhere A2A defines the conversation between two agents, AGNTCY aims at the overall infrastructure (directory, identity, supervision) for multi-agent systems at scale. For a multi-brand luxury group, it is the foundation that would let dozens of agents find one another, trust one another and be supervised from a single dashboard, rather than as silos by house.

  12. AI agent

    An AI program capable of acting autonomously: browsing, comparing, purchasing, booking. Goes beyond the chatbot by carrying out complete tasks.

    Luxury contextAn agent can buy a bag for a client without that client ever visiting the website. This radically changes the luxury purchase journey.

  13. AI Citation Rate

    A founding GEO metric: the percentage of relevant queries in which an AI engine cites the brand as a source, with a link to its content. A citation (the AI draws on the content and sources it) is not a mere mention. It carries more weight: the click-through rate from a linked citation is estimated to be four to six times higher than that of a mention without a link.

    Luxury contextThe north star of a generative-visibility programme: how often ChatGPT, Perplexity, Gemini or Claude draw on the house as a source. Combined with share of voice, it distinguishes absolute presence from relative presence. Caveat: linked citations are visible only on engines that display their sources (Perplexity, AI Overviews).

  14. AI Fashion Week

    A global competition co-founded in 2023 by Maison Meta, in which designers present collections (15 to 30 silhouettes) created with generative-AI tools. Winners join a fashion-tech incubator; their pieces are produced and sold by Revolve Group (Revolve.com and its luxury division Fwrd).

    Luxury contextThe first edition gathered more than 130 candidate collections; a second season followed in 2024 (Milan, New York), with a jury blending Pat McGrath and profiles from Celine and Adidas. For a house, this format prefigures an expanded creative sourcing: testing collections digitally before any industrial investment.

  15. AI Referral Traffic

    Traffic from visitors arriving on a site after a link or recommendation from a generative AI tool (ChatGPT, Perplexity, Gemini, Claude, Copilot). In 2025-2026, a marginal channel (from a few tenths of a percent to around 1% of traffic) but growing at triple-digit rates. ChatGPT accounted for roughly three quarters of it.

    Luxury contextThe stakes lie not in volume but in quality: these visitors are already informed, have strong purchase intent, and convert better than organic traffic (from around +30% to several times higher depending on the source). The house must isolate this channel in its analytics tools to distinguish it from direct or social traffic. Its rise signals greater visibility in AI answers.

  16. AI Scent Advisor (Jo Malone London)

    A conversational agent launched in December 2025 by The Estée Lauder Companies for Jo Malone London, built with Google Cloud's Vertex AI and Gemini. It recreates the in-store fragrance consultation: the customer describes what they are looking for in natural language, and the agent translates it into bespoke recommendations drawn from the brand's olfactory data. Initially in the United States and the United Kingdom.

    Luxury contextA concrete case of a brand assistant that transposes a rare advisory craft (the in-store nose) to digital scale, while keeping the house's vocabulary. The underlying tension — persuading someone to buy a fragrance without smelling it — makes it a textbook case for judging what AI truly augments in a sensory ritual, without diluting it.

  17. AI Share of Voice

    The percentage of a brand's mentions in the answers of AI engines (ChatGPT, Perplexity, Gemini, AI Overviews), measured against its competitors across the same set of queries. It breaks down into mention rate, positioning and comparative share.

    Luxury contextThe metric that turns AI visibility into a competitive reading: is the house cited more or less than its rivals on the queries that matter. According to AthenaHQ's State of AI Search 2026, the average mention rate is only 17.2%, meaning there is ground to win. It helps map the territories where the brand dominates or fades.

  18. AI slop

    A pejorative term for low-quality digital content mass-produced by generative AI, perceived as sloppy and devoid of substance. The phrase spread from 2024 onward (popularised by Simon Willison). In December 2025, Merriam-Webster made slop its word of the year.

    Luxury contextIn December 2025, Valentino faced a torrent of criticism after an AI video for the DeVain bag (around €3,500), even though it was flagged as AI content: the uncanny-valley effect (distorted limbs, dissolving bodies) was deemed disturbing. For a house, even a transparent one, an AI production judged lazy damages the perception of craftsmanship and value, the heart of Luxury.

  19. AI Visibility Score

    An index measuring how reliably generative AI systems find, understand, mention and cite a brand. At its simplest, it is the percentage of responses in which the brand is mentioned; at its most rigorous, a composite index (entity resolution, mention rate, citation rate, source authority, cross-engine consistency).

    Luxury contextIt becomes the dashboard of a house's digital health when the client discovers it through a synthesized response. A low score signals that AI systems are recommending its competitors when a client describes a need without naming a brand. Caveat: there is no standardized definition of the calculation; each vendor (Profound, LLM Pulse) applies its own formula.

  20. AI-referred traffic / "AI Assistant" channel (GA4)

    Visits arriving at a site after a click from an AI assistant. On 13 May 2026, Google Analytics 4 added a dedicated "AI Assistant" channel, with no configuration required. Google cites ChatGPT, Gemini and Claude. Limitation: any AI traffic without a referrer header (apps, copied links) falls into the "Direct" channel and escapes measurement.

    Luxury contextA house can finally quantify the real share of its visits originating from AI. According to Adobe Analytics (Q1 2026, US retail), AI traffic converts 42% better, generates 37% higher revenue per visit and 48% more time on site. To be handled with caution: the channel structurally underestimates the phenomenon.

  21. ALCHEMI (NVIDIA)

    NVIDIA platform for molecular simulation at the atomic scale, integrated by L'Oréal for beauty R&D. A hundredfold gain on molecular-discovery time.

    Luxury contextThe barrier to entry in luxury beauty becomes molecular: whoever simulates faster formulates first. L'Oréal uses it for photoprotection and skin-tone modulation.

  22. Android XR

    Google operating system dedicated to extended-reality headsets and glasses (XR = Extended Reality, encompassing VR, AR and MR). A competitor to Apple's visionOS. Announced in late 2024, with the first consumer products expected in late 2026.

    Luxury contextThe technology platform chosen for the Gucci × Google smart glasses, confirmed by Luca de Meo for 2027 (on the sidelines of the Florence CMD). Same foundation as Project Aura (the first Android XR consumer glasses). Kering is choosing to arrive one cycle after the mainstream. Premium positioning above Ray-Ban Meta (EssilorLuxottica/Meta).

  23. AP2 (Agent Payments Protocol)

    An open standard launched by Google on 16 September 2025, backed by more than 60 partners (Mastercard, PayPal, American Express, Adyen). It allows an AI agent to pay on a client's behalf through cryptographically signed "mandates": intent (scope and budget), cart (items and prices), payment. It is neutral as to method: card, bank transfer or stablecoin.

    Luxury contextFor a luxury house, AP2 makes an agent-driven purchase enforceable and traceable, and therefore defensible in the event of a dispute: a signed chain of evidence attests to consent, which reduces disputes and chargebacks on high-value baskets. Still an emerging standard, to be watched as a promise of architecture rather than installed practice.

B

3 terms

  1. Beauty Genius (L'Oréal)

    L'Oréal's AI beauty agent that orchestrates personalised skincare routines. Rolled out from the website to WhatsApp, entering the customer's private messaging.

    Luxury contextThe AI agent no longer stays on the house's website. It enters private messaging, orchestrating a physical skincare routine from the most intimate space of daily life.

  2. Brand-authored interpretation layer

    An agentic experience designed and steered by the brand, which translates a client's natural-language intent into a selection faithful to the house's identity. McKinsey distinguishes it from a mere agent: the house supplies the machine-readable signals (savoir-faire, provenance, rarity, style); failing that, the agent optimizes what it measures (price, availability, popularity) and flattens brand value.

    Luxury contextThe first interpretation of a client's desire now often takes shape within a general-purpose assistant (ChatGPT, Gemini). By building its own layer, the house reclaims authority over this first filter: it is the house that decides how "a bag for a summer wedding" becomes a recommendation. McKinsey: agentic in luxury is not automation, it is mastered interpretation. Most houses are still at the pilot stage.

  3. BtoA2C (Business to Agent to Consumer) — LUXE ÆTERNAI concept

    The new commerce model in which the AI agent inserts itself between the brand and the consumer. It reformulates the recommendation, compares offers, narrows the shelf from 50 products to 2-3 suggestions, and sometimes buys on the customer's behalf.

    Luxury contextThe traditional BtoC journey does not disappear; it is now preceded by a layer the brand does not yet control.

C

13 terms

  1. C2PA (Coalition for Content Provenance and Authenticity)

    An open standard for the cryptographic provenance of digital content, co-founded by Adobe, Microsoft, BBC and Intel. It attaches signed metadata to media (image, video, audio) documenting their origin, their edits and any presence of AI.

    Luxury contextThe preferred mechanism for meeting EU AI Act Article 50. It makes disclosure invisible to the eye (metadata), readable by platforms, and verifiable by consumers who have the tool. A possible compromise between legal compliance and minimizing the Semiotic Tax.

  2. Capex (Capital Expenditure)

    Industrial investment spending: acquisitions of tangible assets (boutiques, factories, production tools) and intangible assets (software, IT platforms, AI infrastructure) intended to be used over more than one financial year.

    Luxury contextReconKering target: Capex held at 5-6% of revenue. An envelope that must cover network rationalization (Gucci 600 → 450 boutiques), Group-wide AI platform investment, and product projects (Gucci × Google, HModa) all at once.

  3. CDO (Chief Digital Officer)

    The digital director. In luxury, leads digital transformation, e-commerce and, increasingly, AI strategy.

    Luxury contextThis is LUXE ÆTERNAI's target reader. The person who must arbitrate between AI innovation and the preservation of the house's codes.

  4. Chapter 11

    A court-supervised protection procedure in the United States that allows a financially distressed company to restructure while continuing to operate, under the oversight of a federal court.

    Luxury contextSaks Global filed for it in January 2026 with $4.7 billion in debt, an illustration that the best AI stack does not make up for a broken business model.

  5. ChatGPT Atlas

    An AI-powered web browser launched by OpenAI on October 21, 2025 (initially macOS), built on Chromium. It embeds ChatGPT in a sidebar that summarizes and answers questions about the page. A paid agent mode takes over to browse, fill in forms, and carry out tasks (booking, creating a document).

    Luxury contextAtlas shifts the client's point of entry: product discovery begins in a browser where an agent filters, compares, and summarizes, rather than on the brand's website. The house's product page, visuals, and storytelling must remain legible and usable by this intermediary, or risk being bypassed before the official site.

  6. Claude for Chrome

    Anthropic's browser agent, announced on 25 August 2025 as a Chrome extension and progressively expanded in beta in late 2025. In a side panel, Claude reads the page, clicks, fills in forms and chains tasks together, with site-by-site permissions. It requires confirmation before any high-risk action (purchase, data sharing) and blocks sensitive sites (finance, adult, pirated) by default.

    Luxury contextThe default blocking of finance and the confirmation before purchase show that agent developers draw their own red lines, independently of brands. For a house, the risk is to see its high-end journeys reduced to mere transactions to be confirmed, losing the staging that gives value to the act of purchase when a third-party agent operates its site.

  7. Clienteling

    The personalised relationship between a sales advisor and their high-value clients. Historically founded on human knowledge, now augmented by AI.

    Luxury contextClienteling is at the heart of the in-store luxury experience. AI augments it (data, prediction, timing) without replacing the human relationship.

  8. Code of Practice on the Transparency of AI-Generated Content

    A voluntary operational framework published by the European AI Office on 10 June 2026 to meet the transparency obligations of Article 50 of the AI Act. Two strands: on the provider side, machine-readable marking of AI-generated audio, image, video and text content; on the deployer side, the labelling of deepfakes and AI-generated text. Article 50 becomes applicable on 2 August 2026.

    Luxury contextA house that publishes a campaign visual created or retouched by AI, or an automated text, will have to make it identifiable: machine-readable marking on the production side, visible labelling for a deepfake on the distribution side. Article 50(4) provides an exception for works that are "manifestly artistic, creative, satirical or fictional", limiting the obligation to a discreet disclosure: leeway for campaigns openly assumed as such.

  9. Comet (Perplexity)

    Perplexity's agentic web browser, built on Chromium, launched on July 9, 2025 (Windows, macOS) and then made free worldwide in October 2025. Its sidebar assistant summarizes a page, compares products and prices across sites, fills in forms, and completes simple transactions. A background feature (Max plan) runs several tasks in parallel.

    Luxury contextComet illustrates the purchase decided within an assistant that compares offers on the client's behalf, including on price and availability. For a house, the experiential value (advice, staging, scarcity) must be expressed in data the assistant can read, failing which the product is reduced to a line in a comparison against competitors.

  10. Common European Data Spaces

    Sector-specific infrastructures led by the European Commission for the secure sharing of data. Fourteen spaces are in development (health, industry, culture, energy, mobility, finance, etc.). Article 33 of the Data Act imposes interoperability requirements on participants.

    Luxury contextThere is no "fashion" or "commerce" space to date: traceability in Luxury runs through the Digital Product Passport (DPP) arising from the ESPR, whose textile component is expected around 2027. The stake is interoperability: mastering these standards means keeping control over the traceability of one's heritage.

  11. Compliance as Infrastructure

    Regulation (EU AI Act, DPP) is not merely a constraint: it is an infrastructure of trust. The regulator does not slow the agent down, it equips it.

    Luxury contextThe DPP is not a form to fill in; it is the passport the agent will present before recommending the product.

  12. Cost per Resolution

    An amount billed only when an AI agent resolves a request end to end, without human handoff, as opposed to cost per interaction. In 2026, it ranges from roughly $0.50 to $2.00 ($0.99 at Fin AI/Intercom, $0.50 at HubSpot) versus ~$7.40 for a human agent (McKinsey). Each vendor defines "resolution" in its own way.

    Luxury contextThe metric aligns cost with the service delivered, not with the volume of exchanges. For luxury, the trap is to aim for gross savings: a resolution rate pushed too far sacrifices the high-end relationship expected by a UHNWI client. The true reading is the cost per resolution acceptable given the level of service required.

  13. Crystallisation (luxury marketing)

    Stendhal's metaphor (1822) applied to luxury: desire forms in absence and anticipation, not in ease of access. Friction is the product.

    Luxury contextEvery obstacle (waiting list, selection, scarcity) is a layer of crystals. The AI agent removes these obstacles — a fundamental tension.

D

10 terms

  1. Data Act (European data regulation)

    European regulation (EU 2023/2854), applicable since 12 September 2025, governing access to and the sharing of data generated by connected objects. It gives users control over this data, distributes its value more fairly, and removes the barriers to switching cloud providers. Subordinate to the GDPR.

    Luxury contextA House that equips its products with chips (NFC, RFID) or offers connected services will have to give clients access to the data they generate and facilitate its portability. This reshuffles the cards of a clienteling model built on the exclusive capture of client data.

  2. Data obesity

    The accumulation of data without governance or purpose, by analogy with information obesity (David Shenk, 1993). Too much data, everywhere, in formats no one reconciles.

    Luxury contextHouses do not lack data. They have too much of it: siloed CRM, fragmented purchase histories, unstructured product data. LVMH (centralising 75 houses) is the exception. The others are in a state of data obesity.

  3. Data sovereignty (Sovereign AI)

    A house's ability to retain exclusive control over its data and AI models: hosting, access, jurisdiction. It relies on in-house deployments or a sovereign cloud, encryption, and strict partitioning. A geopolitical concern in light of the US Cloud Act (2018). At LVMH, Franck Le Moal sums it up: "technology has become a matter of sovereignty"; the group partitions each brand and has built its internal assistant, MaIA.

    Luxury contextHigh-end client data and creative archives are both a rare asset and a risk. Sovereignty makes it possible to train tools on this heritage without entrusting it to a third party or exposing it to a foreign jurisdiction, and it anticipates GDPR and the AI Act. In April 2026, the European Commission selected four players (including OVHcloud and Scaleway) for a 180-million-euro sovereign-cloud framework.

  4. Digital body language

    Reading the behavioural micro-signals of anonymous visitors (clicks, scrolls, dwell time) without cookies or declared data, to personalise in real time.

    Luxury contextDeployed by Kahoona for Dior. Activates 96% of anonymous visitors usually lost. +24.2% clicks, +12.8% average basket.

  5. Digital Product Creation (DPC) / digital sampling

    The design of a product entirely in 3D: CAD patterns and digitised materials are assembled in simulation software (CLO3D, Browzwear) to obtain a digital twin of the garment before any physical production. Digital sampling replaces physical prototypes with these 3D models validated on screen.

    Luxury contextThe approach sharply reduces the number of physical samples. Hugo Boss reports more than 30% fewer samples and up to 85% time savings from design to shop floor; Tommy Hilfiger claims up to 80% fewer prototypes. Benefits: less waste, shorter cycles, and a virtual showroom following the principle of sell first, produce later.

  6. Digital shelf

    The full set of products visible online to a consumer. It is compressing under the effect of AI agents that pre-filter recommendations (from 50+ references on the shelf to 1-3 on ChatGPT).

    Luxury contextWhen the AI agent recommends only 2 or 3 products, brand visibility depends on optimisation for AI engines (GEO), not on classic visual merchandising.

  7. Digital Twin Factory

    A system for producing marketing content from 3D digital twins of products. Each physical object is digitized (in OpenUSD format) and can be turned into visuals at industrial scale.

    Luxury contextMoët Hennessy generated 3 million visuals across 70 markets via a Digital Twin Factory (Grip/NVIDIA Omniverse). It solves the logistical nightmare of producing multi-market content while maintaining brand consistency.

  8. DPP (Digital Product Passport)

    Mandatory digital identity record (ESPR regulation, expected by 2027) attached to every physical product: materials, origin, environmental footprint, manufacturing conditions.

    Luxury contextThe DPP is not a form to fill in: it is the passport the AI agent will present before recommending the product. The Aura Blockchain Consortium (LVMH, Prada, Richemont) has already registered 70 million products.

  9. Dual readership

    The requirement for any luxury content to be readable both by the human (emotion, desire, brand narrative) and by the AI agent (structure, data, schema.org).

    Luxury contextA purely emotional text is unreadable by machines. A purely structured text is arid for humans. The two readings are not contradictory, but they are not automatically aligned either.

  10. Durable Content Credentials

    A strengthening of the C2PA standard that makes a content item's provenance resistant to the manipulations that strip its metadata (screenshots, recompressions, format changes). It combines cryptographically signed metadata, an invisible watermark, and a perceptual fingerprint, the latter two (soft binding) making it possible to recover the C2PA manifest even after local deletion. The C2PA 2.3 specification added support for live video streaming.

    Luxury contextConventional provenance certificates break the moment an image passes through social media or a messaging app, which re-encode files and erase metadata. For a house seeking to prove that a campaign photo is authentic and not an AI fake, the durable version ensures the proof of origin survives sharing. A direct lever against visual counterfeiting, brand impersonation, and advertising deepfakes.

E

5 terms

  1. EBITDA

    Earnings Before Interest, Taxes, Depreciation and Amortization. A measure of operating profitability and a standard in luxury financial analysis.

    Luxury contextHermès posts a record EBITDA above 40%. It is the indicator that separates the houses investing in AI out of conviction from those doing so out of necessity.

  2. Emotional delegation ceiling

    The threshold beyond which the customer refuses to delegate the purchasing decision to an AI agent. In luxury: level 2 (the agent recommends, the human decides).

    Luxury contextThe ceiling is not technical but emotional. Luxury rests on desire and conscious decision. Source: McKinsey, March 2026.

  3. ESPR (Ecodesign for Sustainable Products Regulation)

    The European regulation on the ecodesign of sustainable products (EU 2024/1781), in force since 18 July 2024. A framework extending environmental requirements to almost all product categories through delegated acts. It makes the Digital Product Passport mandatory and introduces a ban on destroying unsold textiles, applicable from 19 July 2026 for large companies.

    Luxury contextFor a fashion or leather-goods house, product data (composition, origin, durability, repairability) becomes a structured legal obligation rather than a marketing argument. The ban on destroying unsold goods affects a practice long widespread in Luxury: from 19 July 2026, large houses will have to justify it or give it up. The ESPR is fuelling strong demand for AI tools to govern product data.

  4. EU AI Act Article 50

    Provision of the European AI regulation requiring the marking (watermarks, metadata) of all AI-generated content: audio, image, video, text. Effective 2 August 2026.

    Luxury contextLuxury houses already producing millions of AI-generated assets will have to mark them. Compliance falls between marketing, legal and tech, and in luxury, whatever falls between the silos is owned by no one.

  5. Experiential scarcity

    Scarcity based not on owning a limited object but on access to a unique, hard-to-obtain experience: drops, made-to-measure, invitation-only events. It answers the shift from having to living (Millennials, Gen Z) and lasts longer than a product, which by nature can be replicated or resold.

    Luxury contextIn the age of agents, where AI absorbs comparison and functional decision-making, a house differentiates itself on what no interface can replicate: immersion, emotion, human connection. McKinsey notes that desire is often interpreted outside the brand, within generalist assistants. A private dinner, an atelier visit, early access become an asset that agents can neither standardize nor resell.

F

1 term

  1. Frontier model

    A general-purpose AI model among the most advanced of the moment, trained with considerable computing resources, capable of surpassing the state of the art across multiple domains (reasoning, code, multimodality, agentic tasks). Distinct from the "foundation model", a broader term. Examples for 2025-2026: GPT-5, Claude Opus 4.5, Gemini 3 Pro.

    Luxury contextChoosing the right model determines the quality of a house's agents and assistants: a gap in capability translates into the relevance of clienteling and the reliability of responses. A nuance for a decision-maker: "frontier" does not mean "the biggest"; the right trade-off is the model suited to the task and to the House's requirements, not the most powerful.

G

4 terms

  1. GenAI creative agency (e.g. Maison Meta)

    A specialised creative studio that produces campaigns and visuals for fashion, beauty and Luxury by training bespoke generative-AI models tuned to a brand's aesthetic. Maison Meta, founded in 2022 by Cyril Foiret, claims to be the sector's first Gen AI studio (clients: Moncler, Dolce & Gabbana, L'Oréal, Zara) and launched AI Fashion Week.

    Luxury contextThis model makes it possible to generate display-ready visuals without a photoshoot and to turn a technical sketch into a photorealistic product visual. The central stake for a house is the protection of identity and intellectual property: Maison Meta emphasises local workflows (zero cloud) to prevent any dispersion of brand assets.

  2. GEO (Generative Engine Optimization)

    Optimizing a brand's visibility in the answers of generative search engines (ChatGPT, Gemini, Perplexity, Claude). Distinct from SEO: the signals are different. A market worth $200M+ in 2026.

    Luxury contextThe gateway to the internet is changing. Houses that do not optimize for AI agents are invisible in the new customer journey.

  3. GMV (Gross Merchandise Value)

    The gross value of merchandise sold on a platform. A key indicator of transaction volume.

    Luxury contextOn luxury resale platforms (The RealReal, Vestiaire Collective), GMV measures the real appetite of the second-hand market.

  4. GPAI Code of Practice

    A voluntary tool published by the European AI Office on 10 July 2025 to help providers of general-purpose AI models demonstrate compliance with the AI Act. Three chapters: Transparency, Copyright, and Safety and Security. The underlying obligations (Articles 53 and 55) have applied since 2 August 2025. Signatories: Anthropic, Google, Microsoft, OpenAI, Mistral AI.

    Luxury contextA house that relies on generative models (text, visuals, sales assistants) stands to gain from knowing which of its providers have signed this code: it is a signal of compliance and documented governance, particularly regarding the origin of training data and copyright. A concrete criterion to factor into the choice of a technology partner, beyond model performance alone.

H

2 terms

  1. Hallucination (AI)

    A response generated by an LLM that seems plausible but is factually wrong. A major risk in critical applications.

    Luxury contextAn AI that invents a price, an availability or a house's history is a major reputational risk in luxury.

  2. Hyper-personalization

    Real-time personalization, driven by data and AI, that anticipates a client's needs before they are even expressed. It goes beyond "Hello [first name]": it curates recommendations, identifies the right moment, and adapts every touchpoint. AI delivers at scale an attention once reserved for top clients, but it works only when grounded in clean, well-governed proprietary data.

    Luxury contextFor a house, it turns every interaction into a loyalty lever and creates an emotional switching cost: a client who feels understood stays, almost regardless of price. Two downsides: maintaining a consistent brand voice and not crossing the threshold into intrusion (nearly one client in two finds a targeted solicitation right after a visit unsettling). An emerging counter-move in 2025: "contextual" personalization, which predicts preference from context rather than accumulating data.

I

2 terms

  1. Intent Mandate (AP2)

    The first of the mandates in the AP2 protocol (Google, September 2025): a signed digital contract that captures the client's request upstream of the cart. In "human-absent" mode, the client signs it in advance and sets the agent's rules in it (price ceiling, conditions, time window), serving as pre-authorised proof in the sequence intent → cart → payment.

    Luxury contextThis is the term that materialises the shift from the present client to the delegating client. A house may allow an agent to buy a limited edition the moment it goes on sale, but only within the limits signed in advance (price ceiling, a single piece). It becomes the contractual safeguard for a purchase that no one validated in real time.

  2. Inverted Authenticity

    In a world where AI produces perfect imitation, the real can no longer be taken for granted — it has to be proven. Authenticity used to be an inheritance; it is becoming a technology.

    Luxury contextThose who reject the agent in the name of the real deprive themselves of the only tool capable of proving it at scale. Trust in the physical object now depends on the machine that certifies it.

K

1 term

  1. KPI (Key Performance Indicator)

    Key performance indicator. A metric used to measure the achievement of strategic or operational objectives.

    Luxury contextLuxury KPIs (NPS, lifetime value, average basket) must be rethought in the agentic era. What do you measure when it is an agent that buys?

L

3 terms

  1. Levels of agentic autonomy (AEPD framework)

    A gradation set out by the Spanish data protection authority (AEPD, February 2026 guidance): the controller must define the agent's degree of autonomy, from "the agent proposes, the human executes" to "the agent executes, the human observes." The higher the autonomy, the more stringent the GDPR requirements. Legal liability remains with the data controller.

    Luxury contextWhen a House deploys an agent capable of acting on its own (responding, modifying an order, closing a transaction), this framework provides the vocabulary to set the "delegation ceiling" and to place human checkpoints. Directly applicable for framing a clienteling assistant before it goes into production.

  2. LLM (Large Language Model)

    A large-scale language model, trained on massive text corpora. The technical foundation of generative AI (GPT, Claude, Gemini).

    Luxury contextIt is the engine that powers AI agents, chatbots and personalization tools. Understanding its limits prevents missteps.

  3. llms.txt

    A Markdown file placed at the root of a website, proposed on 3 September 2024 by Jeremy Howard (Answer.AI), which offers large language models a selective map of the content deemed most important. It remains a community proposal, not an official standard.

    Luxury contextFor a house, a lever of narrative control: designating the canonical pages (history, craftsmanship, collections) meant to feed the answers. Major caveat in 2026: no major player (OpenAI, Google, Anthropic, Meta, Mistral) uses it in production, with Google having stated that it does not support it. A bet on what's coming, not a guarantee.

M

8 terms

  1. Machine customers

    Non-human economic actors that obtain goods or services in exchange for payment (Gartner's definition): voice assistants, connected objects, procurement systems or AI agents that decide and trigger the purchase. Gartner suggests that 20% of customer service contacts will come from them by 2026, and up to 20% of revenue by 2030.

    Luxury contextA luxury house must decide whether it is willing to sell to a non-human buyer. Two stakes: emotional friction (selling to a machine becomes complicated when emotion and status dominate, the core of the luxury purchase) and disintermediation (going through a third-party agent means losing the relationship and the client's data). Illustrated by Kering's KNXT and its assistant Madeline.

  2. Machine-readable

    A brand's ability to have its data read and interpreted by AI agents. Whatever is not documented for LLMs does not surface in recommendations. The llms.txt format is an emerging standard.

    Luxury contextHouses whose product pages, history and values are not structured for LLMs lose agentic visibility.

  3. MaIA (LVMH)

    LVMH's internal AI platform, deployed across the group's 75 houses for trend prediction, logistics optimisation and personalisation.

    Luxury contextThe first group-wide AI deployment in luxury. A sector benchmark for gauging the AI maturity of competitors.

  4. Mastercard Verifiable Intent

    An open framework announced by Mastercard on 5 March 2026 with Google, producing tamper-proof cryptographic proof of the authorisation given by a customer delegating a purchase to an agent. It records in a single entry the customer's identity, their instructions and the transaction outcome, forming an audit trail that can be consulted in the event of a dispute. Designed to interoperate with Google's AP2 protocol.

    Luxury contextThis is the link that turns customer consent into enforceable proof, and thus into a compliance element. For a luxury house, the value lies in traceability: on a purchase delegated to an agent, it can demonstrate the alignment between the customer's intent, the agent's action and the debit, which reduces disputes on high-stakes transactions.

  5. MCP (Model Context Protocol)

    Protocol developed by Anthropic to connect LLMs to external data sources and tools in a standardised way.

    Luxury contextIt lets an AI agent access a house's CRM, inventory or catalogue in real time. An essential building block of augmented clienteling.

  6. MCP Registry

    An open, official catalogue of MCP servers (the connectors linking an AI agent to a company's tools and data), launched in preview on 8 September 2025 by the Model Context Protocol working group (Anthropic, GitHub, Block, PulseMCP). A single reference source for discovering public connectors, with a verified-naming system (DNS, HTTP or GitHub account) that prevents impersonation.

    Luxury contextA trusted directory of the connections between agents and business systems. For a house, the challenge is software supply-chain security: ensuring that a connector to its customer-management or inventory tool is genuinely official and not a counterfeit, the way one authenticates a supplier.

  7. Mention Position

    A metric measuring where a brand appears in an AI-generated response: at the start, in the middle or at the end. The first position concentrates the strongest engagement; being relegated to fifth position or beyond across the majority of responses signals a visibility problem. The KPI is tracked by topic, by model and by competitor.

    Luxury contextFor a house, being cited is not enough: being named at the top of the response of an agent comparing bags or watches carries far more weight than a passing mention. It is the agentic equivalent of the end-of-aisle display, transposed into the text the AI recites to the client — an indicator of desirability in the language of the models.

  8. Model digital replica

    A digital representation of a model's image or voice, generated by AI, that reproduces or replaces their appearance or performance. The New York Fashion Workers Act, in force since 19 June 2025, requires explicit written consent, separate from the contract (scope, purpose, duration, remuneration). Routine retouching is excluded from it.

    Luxury contextA house that creates an AI model from a real person, or reuses their image beyond the initial shoot, must secure this detailed consent. The law provides for civil fines and a right of action for the model (damages increased by up to 300% in the event of wilful misconduct). The chain of image rights becomes a legal control point ahead of any AI campaign.

N

3 terms

  1. Nano Banana Pro (Gemini 3 Pro Image)

    An image generation and editing model from Google DeepMind, launched on 20 November 2025 and derived from Gemini 3 Pro. It blends up to 14 images while preserving the appearance of 5 people, allows localized edits, and renders legible, typeset text within the image, in 2K and 4K.

    Luxury contextA house can render the same model or the same bag across dozens of scenes without a new photo shoot, with a consistent visual identity. Useful for lookbooks and e-commerce visuals, provided the aesthetic standard of Luxury is upheld. All images carry an invisible SynthID watermark.

  2. NemoClaw

    Nvidia's open-source enterprise AI agent platform, unveiled at GTC (GPU Technology Conference) in March 2026. A full agentic stack for deploying autonomous agents in the enterprise.

    Luxury contextAgentic infrastructure is becoming widely available. The tools are here. What luxury lacks is the organisational will to adopt them.

  3. NLWeb

    An open-source framework unveiled by Microsoft at its Build conference in May 2025, designed by R.V. Guha (architect of RSS and Schema.org), which turns a website into a conversational interface that can be queried by humans and AI agents alike. It builds on existing semantic markup (Schema.org, RSS) and exposes each site as an MCP server, queryable in natural language with structured JSON responses.

    Luxury contextNLWeb makes a house's website directly readable by agents, with no heavy overhaul: a client's agent can ask about the availability of a model or the details of a collection and obtain a reliable answer, drawn from the official source. A lever of agentic visibility that extends the investment already made in structured data. Early adopters: Shopify, Tripadvisor, Snowflake.

O

2 terms

  1. OpenUSD

    Universal Scene Description: an open standard for describing 3D scenes, developed by Pixar and championed by NVIDIA. It enables interoperability between 3D creation tools.

    Luxury contextFormat used by Moët Hennessy to digitise its bottles into 3D digital twins on NVIDIA Omniverse. It allows the generation of marketing visuals for multiple markets from a single 3D asset.

  2. OpenUSD 3D product configurator

    A product-personalisation application built on OpenUSD (the open 3D scene-description format), NVIDIA RTX rendering and generative AI. It displays a photorealistic, physically accurate digital twin that the customer configures in real time (colour, material, finishes), with possible visualisation in extended reality (Apple Vision Pro headset).

    Luxury contextFor a Luxury house, such a configurator offers a personalised online purchasing experience for high-value items (leather goods, watchmaking, automotive), faithful to the real material. NVIDIA documents gains on the order of 40% time savings in content production. Studios such as Katana (used by Nissan) or Media.Monks build these pipelines.

P

4 terms

  1. Phygital

    A blend of "physical" and "digital" that emerged in the early 2010s. Phygital refers to the fusion of physical and digital spaces into a seamless purchase experience, where every touchpoint (social network, website, augmented reality, app, store) is connected to the others. In luxury, it aims for a fluid, emotional experience in which data and AI feed personalization without erasing the human relationship.

    Luxury contextA client may discover a product on Instagram, explore it in augmented reality, then enter a store where the advisor already knows their journey, and complete the purchase on mobile while there. Sephora illustrates the mass-market approach (interactive kiosks, virtual try-ons linked to the app); luxury houses transpose this continuity while keeping exclusivity and the human gesture at the center. Without proof, the word slides into a buzzword.

  2. Physical AI

    AI that operates in the physical world through sensors, molecular simulation and robotics. Term popularised by Jensen Huang (NVIDIA CEO) at GTC in March 2026.

    Luxury contextPhysical AI is arriving because luxury has always been a matter of material. The AI agent discovers that it has a body: no hands, no nose, but it reads a molecule at the atomic scale.

  3. Pilot Purgatory

    The zone where organisations remain stuck between a successful AI pilot and a full-scale deployment. 80% of pilots meet their technical objectives, yet only 23% generate measurable financial impact.

    Luxury contextLuxury is particularly exposed to it: a culture of secrecy, organisational silos and undocumented craftsmanship prevent the move from pilot to deployment, even when the technical KPIs are in the green.

  4. Prompt Volume

    A metric estimating how frequently a keyword, brand or theme appears in natural-language queries put to AI assistants (ChatGPT, Claude, Gemini, Perplexity). The equivalent of search volume for the conversational era. Since the platforms do not publish their data, the figures are modelled from anonymised panels and extrapolated: an indicative estimate, not verifiable data.

    Luxury contextFor a house, it reveals the real conversational demand around its universe: how many clients ask an AI for an investment handbag or a niche fragrance, and in what wording. Combined with brand visibility, it helps prioritise the territories where demand is strong but presence is weak. Caveat: modelled volumes, to be used to spot gaps, not to ground a decision on their own.

Q

1 term

  1. Quiet AI

    A concept theorised by Ekimetrics (2025): a discreet AI that optimises without breaking the magic. Inventory, pricing, demand forecasting, with no interface visible to the customer.

    Luxury contextThe common thread among AI deployments that scale in luxury hospitality: the AI agent remains invisible. The technology disappears behind the emotion.

R

2 terms

  1. R&I (Research and Innovation)

    The research and innovation department of a cosmetics group. At L'Oréal, it encompasses formulation, testing, molecular simulation and the development of new technologies.

    Luxury contextThe integration of AI into L'Oréal's R&I (ALCHEMI) transforms the barrier to entry: it becomes molecular, not marketing.

  2. ROCE (Return On Capital Employed)

    Return on capital employed. The ratio of operating profit to capital employed (shareholders' equity plus net financial debt). It measures the profitability of the capital committed to the business, independent of financing structure.

    Luxury contextStated ReconKering 2030 target: ROCE above 20%. A preferred indicator for assessing the operational efficiency of a multi-house group with shared infrastructure (HModa joint venture, Group AI platform, consolidated procurement).

S

3 terms

  1. Share of Model (Jellyfish)

    A proprietary metric and platform (Share of Model™) launched by Jellyfish (Brandtech group) in December 2024, which measures how large models (ChatGPT, Gemini, Llama) perceive a brand. It is the equivalent of market share for generative AI: the proportion of mentions or recommendations obtained within a category.

    Luxury contextAsked about style, the AI returns the same cluster of mega-brands (Chanel, Louis Vuitton, Hermès, Gucci, Prada), with independent labels appearing in fewer than 3% of fashion responses. Tracking its Share of Model tells whether a house exists in this space, model by model. Chivas Brothers and Danone tested it in beta.

  2. Shared Payment Token (SPT)

    A payment instrument launched on 29 September 2025 with ChatGPT's Instant Checkout (the Agentic Commerce Protocol from Stripe and OpenAI). Issued by Stripe, it allows an application such as ChatGPT to trigger a payment without exposing the buyer's banking credentials to the merchant. Each token is single-use, bound to a single merchant, an amount and an expiry date.

    Luxury contextA house selling through a conversational agent collects payment without ever seeing or storing the client's card data, the token being valid only for that transaction. This reduces the risk surface, but the house must weigh what it gives up: it is the agent, not the house, that holds the point of payment contact, and therefore part of the client relationship.

  3. SynthID

    Google DeepMind's invisible digital watermark for marking AI-generated content and verifying its origin. Launched in 2023 for images, it now covers image, video, audio, and text, and withstands cropping, filters, and compression. A dedicated detector verifies the signature.

    Luxury contextFor a house, this is the tool to sign its own AI-generated visuals and videos and to detect an image circulating improperly under its name. To be combined with the AI Act (Article 50, August 2026) and the C2PA standard. Limitation: SynthID detects only its own signature and covers only Google's models.

T

5 terms

  1. The 5 Es

    Mickaël Tsakiris's editorial framework: Excellence, Exception, Experience, Emotion, Elevation. A reading grid for assessing any AI deployment in luxury.

    Luxury contextEvery AI initiative must reinforce at least one of these invariants. If none is concerned, the technology has no place in a luxury house.

  2. The Antechamber

    The space between the client's intent and the house's storefront, populated by AI agents that the houses do not control. The place where preferences form before the visit.

    Luxury contextThe houses did not build this space, do not control it, and are represented there by third parties (LLMs, agents). Original LUXE ÆTERNAI concept #6.

  3. The Simondon Line

    The boundary each house must draw between what it delegates to the AI agent and what it keeps as an irreducible human relationship. Inspired by Gilbert Simondon (On the Mode of Existence of Technical Objects, 1958).

    Luxury contextThe sales advisor is not there to answer questions — the agent already does that. The advisor is there to be the relationship between the client and the house.

  4. Timegrapher

    Measuring instrument used in watchmaking to verify the accuracy of a mechanical movement by analysing the sound of the escapement. Rolex patented a next-generation neural-network timegrapher (January 2026).

    Luxury contextAI enters the calibre: Rolex's neural-network timegrapher detects defects and counterfeits through acoustic signature, without touching the watchmaker's craft.

  5. Trusted Agent Protocol (Visa)

    An open framework announced by Visa on 14 October 2025 with Cloudflare, enabling a merchant to recognise a legitimate AI agent and distinguish it from a malicious bot. Built on HTTP Message Signatures (RFC 9421) and aligned with Web Bot Auth, it signs the agent's identity while letting the merchant retain visibility on the actual customer. Twelve launch partners (Adyen, Stripe, Shopify, Microsoft, etc.).

    Luxury contextThe concrete risk for a house: that its anti-bot systems mistakenly block a genuine agent shopping for a good customer, and lose the sale. The protocol lets trusted agents through while preserving the trace of the actual customer, a condition for authorising agentic commerce without forgoing customer knowledge, the bedrock of clienteling.

U

2 terms

  1. UHNWI (Ultra High Net Worth Individual)

    An individual whose net worth exceeds 30 million dollars. A strategic client segment for fine watchmaking, haute couture and high jewellery.

    Luxury contextUHNWI data is among the most sensitive in luxury. An AI agent that accesses it touches the core of a house's reputational capital. This is the crux of the Vault Dilemma.

  2. Universal Cart (Google)

    A unified cart unveiled by Google at I/O 2026, working across merchants and services: items can be added from Search, Gemini, YouTube or Gmail. Powered by Gemini, it tracks prices and alerts when items are back in stock. Payment runs through the Universal Commerce Protocol (UCP), where the brand remains merchant of record, complemented by AP2. Roll-out in the United States from summer 2026.

    Luxury contextUniversal Cart aggregates the purchase decision away from the brand's site. The presence of Sephora and Ulta Beauty shows that prestige beauty is already exposed there to a price-oriented cross-merchant cart logic. A point of vigilance: the UCP's "merchant of record" status makes it possible to remain the registered seller and to retain the customer relationship and data.

V

1 term

  1. Visa Intelligent Commerce

    A programme and API suite launched by Visa in 2025 to make payments initiated by an AI agent acting on a client's behalf reliable: tokenised cards, payment passkeys, configurable spending controls and authentication distinguishing a legitimate agent from a malicious bot (Trusted Agent Protocol, October 2025). In December 2025, Visa announced hundreds of agentic transactions in production.

    Luxury contextFor a luxury house, the assurance that agent-driven purchases travel along the same rails and guarantees as conventional payments, without reinventing its payment funnel. The value lies in the fine-grained rules: capping what an agent commits without fresh validation, restricting categories, tracing every transaction. The house retains control of the client relationship.

W

1 term

  1. Wabi-sabi

    A Japanese aesthetic concept that prizes imperfection, impermanence and incompleteness as sources of beauty.

    Luxury contextA philosophical argument against algorithmic perfection. Controlled imperfection is a marker of craftsmanship and luxury that AI cannot reproduce.

X

1 term

  1. x402

    An open payment protocol launched by Coinbase in May 2025, which revives the HTTP 402 status code to enable settlements in stablecoins (USDC) directly within a web exchange, verified and settled in a few seconds. In April 2026, Coinbase handed it over to a foundation hosted by the Linux Foundation, co-founded with Cloudflare (Google, Visa, AWS, Stripe, Mastercard among the participants).

    Luxury contextx402 paves the way for machine-to-machine micropayments without a card or an account: an agent can pay per use for a service or a piece of data. For Luxury, the issue is the backstage economy, for example an agent that automatically pays, item by item, for an authentication, origin-certification or second-hand data service.

This vocabulary shifts every week. To follow it, the newsletter.

Every LUXE ÆTERNAI edition introduces one or two new concepts and sharpens the older ones. Want to be notified? Sign up at the bottom of the Insights page.

Read the newsletter → See the Insights →