Luxury and specialty retail has always relied on a kind of intelligence that is hard to scale: the ability to read the moment.
A customer checking an order status does not need the same kind of help as a customer searching for an anniversary gift. One wants reassurance. The other wants guidance. Treating both as the same AI use case is where many retailers go wrong.
The more useful approach is to design different AI agents for different customer moments.
Pandora offers a strong example of why this matters. In its digital customer experience work, two distinct agents were created to support two distinct intents. Clara was designed as a customer service agent, focused on routine, high-volume inquiries such as order status and jewelry care. Gemma was designed as a personal shopping agent, helping customers choose jewelry based on occasion, recipient and budget. Both were built to reflect the warmth and expertise of in-store interactions, but they solve very different problems.
That distinction is not a technical footnote. It is the strategy.
Service moments require reassurance, speed and operational truth
When customers enter a post-purchase service journey, their needs are usually practical and immediate. They want to know where an order is, whether it will arrive on time, how to care for a product or what to do next. In these moments, a luxury or specialty brand is not being judged only on efficiency. It is being judged on confidence. Generic answers create friction. Real answers create trust.
This is where a service agent like Clara becomes valuable. Clara automates routine questions in conversational language, but the real strength is that the experience is connected to the systems that make the interaction useful. By integrating with service and commerce platforms and drawing on Pandora’s order management environment, Clara can provide real-time order updates rather than static responses. That matters because post-purchase reassurance depends on visibility.
For retail leaders, the lesson is clear: a service agent should be designed around resolution. Its job is to reduce effort, absorb repeatable volume and give customers clarity in moments that might otherwise create anxiety. When done well, that also changes the economics of service. Pandora reported 60% autonomous case deflection, allowing service specialists to focus on higher-value interactions, along with a 10% uplift in Net Promoter Score driven by agent-first service.
The point is not simply to automate contacts. It is to protect the brand in the moments when reliability matters most.
Shopping moments require discovery, context and taste
A shopping journey is different. Here, the customer may not have a defined answer in mind. They may be exploring for a birthday, an anniversary, a personal milestone or a budget-conscious gift. The challenge is not resolving a task. It is helping the customer move from uncertainty to confidence.
That is why a shopping agent should not be treated like a service bot with a product feed attached.
Gemma shows what a more useful model can look like. As a personal shopper agent, Gemma helps customers discover jewelry by understanding the occasion, the recipient and the shopper’s budget. It uses commerce data, customer insight and UX research to deliver more tailored recommendations. The experience is closer to guided selling than scripted product search. Instead of pushing customers through a narrow path, it helps narrow choices in a way that feels more like an informed associate.
For luxury and specialty retailers, this is especially important. Many purchases in these categories are emotional, occasion-based and meaning-rich. Customers are not only comparing features. They are choosing something symbolic. A shopping agent therefore needs to do more than retrieve products. It needs to support intent, reduce overwhelm and help customers feel good about the decision.
One brand, different agent roles
Clara and Gemma point to a broader design principle: service agents and shopping agents should share a brand voice, but they should not share the same job description.
A service agent should excel at post-purchase reassurance, routine inquiry handling and real-time visibility into operational data. A shopping agent should excel at guided discovery, gifting support, occasion-based recommendations and budget-aware curation. Both can feel warm and conversational. Both can reflect the brand. But the experience design, data requirements and escalation logic should be different.
This is where many AI programs become too technology-led. Leaders start with the model, the platform or the channel. They should start with customer intent.
A practical framework for retail leaders
For brands deciding how to deploy AI agents, four design questions can help separate real value from generic automation.
Map the customer intent.
Start by identifying the moments that matter most. Is the customer trying to resolve, decide or explore? Resolution moments tend to suit service agents. Decision and discovery moments tend to suit shopping agents. If the intent is unclear, the experience will usually feel unclear too.
Define the data each agent needs.
A service agent needs access to the systems behind the answer: order status, service history, fulfillment signals, returns processes and care information. A shopping agent needs a different data mix: product attributes, budget cues, occasion context, behavioral signals and customer insight. The quality of the interaction depends on the quality and relevance of the data behind it.
Identify where human specialists must stay involved.
Not every moment should be automated end to end. Complex service recovery, emotionally sensitive interactions and high-value consultations still benefit from human judgment and empathy. The goal is not to eliminate specialists. It is to free them from repetitive work so they can focus where they matter most.
Design around brand voice, not scripts.
Pandora’s experience highlights the shift from rigid scripted exchanges to more dynamic, context-aware conversations. That shift matters in luxury and specialty retail because customers can immediately tell when an interaction feels mechanical. Brand voice is not a cosmetic layer added at the end. It is part of how trust, care and expertise are expressed digitally.
Experience design is the differentiator
The strongest AI agents are not defined by novelty. They are defined by fit. They match the customer moment, connect to the right data and support the broader operating model behind the experience.
Pandora’s broader transformation makes this point even stronger. The agent experiences were not created in isolation. They sat on top of work to modernize commerce, improve order management, strengthen omnichannel fulfillment and remove silos across teams. That foundation matters because trusted AI depends on trusted operations. An agent can only reassure a customer about an order if the business can see the order clearly. An agent can only guide discovery well if the underlying commerce and customer data are connected.
For retail leaders, that is the bigger opportunity. AI agents should not be treated as one universal interface. They should be designed as a portfolio of experiences, each built around a specific customer need and connected into a broader retail ecosystem.
In luxury and specialty retail, that portfolio approach can make digital feel less distant and more human. It can reduce friction after purchase, improve confidence during discovery and extend the qualities customers value most in stores into digital channels.
The future is not one agent doing everything. It is the right agent, for the right moment, in a way that feels unmistakably true to the brand.