Dreamforce, Salesforce’s annual conference, took place in San Francisco in September. Among its 2026 announcements were AI agents designed to work across customer data and business tools, answer questions and carry out authorised tasks. Its commerce programme also showed how an agent might guide a customer from product discovery towards purchase. For luxury leaders, the significance lies in what these capabilities could do to a relationship that moves between a search, a brand website, a boutique and a client adviser. Salesforce’s Dreamforce announcements
Clients have already changed their behaviour. In an April 2026 survey by Bain and Comité Colbert, 54% of US respondents, 64% of Chinese respondents and 27% of French respondents said they had used AI during their most recent luxury purchase. Even among clients who bought in a boutique, 47% had used AI somewhere in the journey. These are three surveyed markets, not a universal description of luxury consumers; the differences between them matter. Bain and Comité Colbert, June 2026
Each encounter should reflect the house’s product knowledge and understanding of service. Agentic AI could extend that continuity, provided the house decides where technology may act and where an adviser must interpret.
THE CLIENT’S FIRST QUESTION MAY ARISE ELSEWHERE
A client considering a coat for an upcoming trip may begin with a general-purpose AI assistant. By the time they reach a brand’s website, their search may already have been shaped elsewhere. Bain found that luxury clients who used AI most often turned to it for product or brand research and styling inspiration. That does not establish a desire for autonomous purchasing. Bain and Comité Colbert
Luxury brands therefore need to make their products intelligible wherever that first question is asked. An agent needs dependable information on composition, fit, care, availability and the distinctions between pieces. The richer information also matters. Why does a particular cut belong to the house? What does a material allow the maker to do? Which detail would a knowledgeable adviser mention to help a client choose?
This calls for editorial and merchandising judgement. An external assistant cannot convey much of a house’s point of view from a product name, image and price. Accurate knowledge of craft and use improves the information on which it can draw, even though the platform controls its own answer.
Sephora offers a recent example of a retailer entering a conversation already taking place on another platform. In March 2026, it announced a US pilot of its app in ChatGPT. Customers can seek beauty advice there and, if they choose to connect a Beauty Insider account, receive recommendations informed by their profile. Payment within the app was described as a future development, rather than a feature of the initial pilot. Sephora is bringing its own product knowledge and customer relationship into the exchange. Sephora, March 2026
For a luxury house, the corresponding decision is more selective. Which parts of its expertise should be available in a third-party conversation? At what point should the client enter a space designed by the house, speak with an adviser or see a product in person? Presence on an AI platform is one decision; the path back into a distinctive brand experience is another.
MAKE THE DIGITAL EXPERIENCE RECOGNISABLY THE HOUSE’S OWN
On a brand’s own channels, AI can interpret an open-ended request with greater attention to its collections and creative direction. A client may know the occasion but have no product in mind. Helping them explore possibilities becomes valuable when the suggestions express a recognisable taste.
Ralph Lauren’s Ask Ralph, introduced in 2025, illustrates this approach. Within the brand’s US app, a customer can describe an occasion or ask how to wear an existing item. The tool assembles shoppable Polo Ralph Lauren looks from available inventory and allows the customer to refine the suggestions. A conversational request becomes a complete styling proposition rooted in the brand’s own world. It is a styling assistant, rather than evidence of a fully autonomous luxury purchase journey. Ralph Lauren, September 2025
A newer example comes from Brunello Cucinelli. The house launched an AI-based e-commerce site in January 2026 using its Callimacus platform. It describes a digital boutique designed to interpret a visitor’s intent and guide discovery through a personalised journey consistent with the house’s values. The company said its first months of use increased average time spent on the site, while offering its own assessment of the effect on digital sales. Those early observations are encouraging, but they are the company’s account, not an independent measure of the wider commercial impact. Brunello Cucinelli, July 2026 half-year release
Leaders commissioning a digital agent should ask what this experience would do differently if it belonged to their house. The answer may lie in how it explains a material, builds a silhouette or invites a conversation. The demanding work sits in the choices it makes.
An agent trained to maximise basket size may propose more products at every opportunity. A house built on considered selection may favour fewer options, carefully explained. The manner of making a recommendation becomes part of the brand experience.
GIVE ADVISERS BETTER CONTEXT BEFORE THE CLIENT ARRIVES
The most promising clienteling application may be less visible to the client. An adviser preparing for an appointment often needs to gather information from several places: stated preferences, earlier purchases, pieces the client has viewed, available sizes, repairs in progress and stock held by another boutique. The time spent assembling that picture could instead be spent preparing an informed conversation.
An agent could assemble a briefing and check authorised inventory. For a client requesting a jacket, it could distinguish the requested size and colour from a related piece it merely suggests, identify the boutique holding the jacket and show when stock was last confirmed. This illustrates a possible workflow, not a named maison’s deployment.
The adviser interprets that preparation. A previous purchase does not prove that the client wants a similar piece; a preference may have changed. Good clienteling recognises these limits and asks a better question when the client arrives.
This builds on a point we explored in The Luxury of Being Remembered: a client should not have to repeat the same preferences at every appointment, yet memory must never feel like surveillance. Better CRM gives the adviser useful context. The memorable part remains what the adviser notices, asks and does with that knowledge. Technology can help the maison remember; its people give that memory meaning.
The beauty counter offers an adjacent illustration. In our VivaTech 2026 analysis, we observed how AI-supported diagnostics were changing the order of an in-store consultation: assess a need, then discuss a suitable product. L’Oréal presented Kérastase’s K-SCAN scalp diagnostic, already deployed in more than 9,000 salons, and a Lancôme Cell BioPrint-led longevity routine. These are diagnostic tools, not agentic clienteling systems. Their strategic lesson travels across categories: the value of the data appears when a trained professional can interpret it and make the conversation more precise. Our related article on beauty’s longevity shift develops that consultation-led approach.
Kering’s 2026 strategy proposes a client intelligence platform to support deeper engagement across its houses. Shared tools must help each house deliver an experience that still feels its own. Kering, April 2026
DECIDE WHAT THE AGENT MAY DO
The word agentic becomes meaningful when a system can take an action rather than merely produce an answer. That action might be modest, such as checking stock or preparing an appointment note. It might be more consequential, such as sending a message, changing a client record or placing an order. These should not all carry the same level of authority.
Dreamforce’s commerce announcements include a shopper agent that answers questions and compares products. Salesforce has also announced natural-language search for store associates, scheduled for January 2027. The roadmap indicates possibilities, not deployment in luxury boutiques. Salesforce, Dreamforce ’26 commerce innovations
A house can translate that roadmap into three practical levels of responsibility. An agent can retrieve approved product and stock information or assemble a briefing. It can propose a selection or draft a follow-up for an adviser to review. A personal invitation, a sensitive service recovery conversation or a recommendation involving an important client relationship calls for the adviser’s decision.
These boundaries need to be built into the workflow. If a message requires approval, the adviser should see the proposed text, the reason for the suggestion and the information on which it relies. If stock data is uncertain, the system should say so before anyone makes a promise. If the client has not invited further contact, an apparent sales opportunity should not quietly become automated outreach.
The test for a leader is concrete: Would the client expect the house to make this decision automatically? Could an adviser explain why it was made? What happens when the information is wrong?
MAKE TRUST PART OF THE SERVICE DESIGN
Personalisation asks a client to share information. In luxury, that information may reveal far more than a product preference: an important date, a gift recipient, travel plans, a body measurement or a private reason for a purchase. Connecting information across channels increases the duty to handle it with care.
In a 2026 NRF and IBM study, 83% of respondents expressed at least one concern about privacy, data misuse or unwanted marketing. This covers retail consumers generally, rather than luxury specifically. It helps explain why personalisation can feel helpful in one context and intrusive in another. NRF and IBM, 2026
A client-facing assistant should make its role clear. When personal information shapes a recommendation, the client should be able to understand what has been used and choose whether to provide more. Within the business, permissions should determine which information an agent can access and which actions require human review. Salesforce’s security guidance places responsibility on organisations to configure those rules and approval points. The presence of a technical safeguard alone cannot settle a service decision. Salesforce security guidance, February 2026
Accuracy is equally tangible. A client who receives a confident answer about a rare piece, then discovers that it is unavailable, experiences the error as a failure of the house. Product knowledge, live stock, after-sales conditions and appointment details therefore need defined owners and a route for correction. Trust is built through information that can be checked, permission that can be understood and a person who can take responsibility.
START WITH ONE JOURNEY, THEN JUDGE THE EXPERIENCE
Luxury leaders do not need to redesign every touchpoint at once. A focused pilot could begin with one journey: a client researches a piece online, requests an appointment and visits a boutique. Mapping that sequence reveals where information is lost and where an agent might genuinely help.
The pilot should specify what the agent may retrieve or do, and which decisions belong to an adviser. Product, retail, CRM, service and privacy teams must define those boundaries together. Advisers should test briefings in real appointments and report irrelevant or intrusive suggestions.
Alongside conversion and efficiency, leaders should examine stock accuracy, whether advisers correct the briefings, whether clients repeat information and whether the handover to the boutique feels natural. These observations show where the system helps the relationship and where its design needs work.
The potential of agentic AI in luxury lies in the continuity it could give to a house’s knowledge and care. A client can begin with a question to an assistant, explore a collection online and arrive in a boutique to meet someone prepared to listen. A house will make that journey its own through the quality of its information, the judgement it gives its people and the boundaries it sets for technology.
Marine Moureaux-Gall / MAENA Agency
Featured image credit: HIC Global Solutions, “Complete Guide to Dreamforce 2026”; Dreamforce ’26 branding by Salesforce.
