Luxury and specialty retail has always depended on something that is hard to scale: human warmth.

In the store, great associates do more than answer questions. They read context. They understand intent. They guide without overwhelming. And they turn a transaction into a relationship.

The challenge for digital leaders is not simply how to automate more service or add more product recommendations. It is how to bring the warmth of the store into digital channels in a way that still feels personal, helpful and brand-right.

Pandora’s recent work with Publicis Sapient and Salesforce offers a strong model for how to do exactly that. Together, the teams used Agentforce to create two distinct AI agents: Clara for customer service and Gemma for guided shopping. The goal was not to create scripted bots that push customers down narrow paths. It was to design intelligent digital experiences that could reflect more of the warmth, expertise and conversational quality customers associate with in-store interactions.

That distinction matters. In luxury and specialty retail, customer expectations are high even when the question seems simple. A shopper asking where an order is does not want to feel trapped in a support flow. A gift buyer browsing for jewelry does not want to feel like they are using a search filter with a human name. If digital agents are going to add value, they need to feel aware of customer context, connected to real business systems and capable of helping customers move forward with confidence.

Pandora built Clara and Gemma around those principles.

Clara was designed as a customer service agent focused on high-volume, routine inquiries such as order status and jewelry care questions. It integrates with Service Cloud and Commerce Cloud, and uses Pandora’s IBM Sterling environment to provide real-time order updates. That means Clara is not just responding with generic language. It is drawing on live operational data to help customers resolve common needs quickly and conversationally.

Gemma was designed for a different moment in the journey: discovery and consideration. As a personal shopper agent, Gemma helps customers choose jewelry based on the occasion, the recipient and the customer’s budget. To do that, it draws on data from Commerce Cloud, Bloomreach and Pandora’s UX research through Data Cloud, enabling more tailored recommendations. The experience is closer to guided selling than search. Instead of asking customers to do all the work, Gemma helps narrow choices in a way that feels more like a helpful associate.

The design philosophy behind both agents is as important as the technology stack. Publicis Sapient helped shape the strategy, build the agents and design interactions that moved away from rigid scripts toward more dynamic, context-aware conversations. For retailers, this is the real shift. The value of AI agents is not just in answering faster. It is in making digital interactions feel more adaptive, relevant and human.

That human quality should not be misunderstood as trying to imitate store associates word for word. It means understanding what makes the in-store experience effective in the first place. Usually, it comes down to three things: context, confidence and continuity.

Context means the interaction reflects what the customer is trying to do right now. A post-purchase customer needs quick resolution, not inspiration. A gifting customer may need help thinking through recipient, style and price point. Confidence means the response is grounded in real data and useful recommendations, not generic answers. Continuity means the experience feels connected to the broader journey, including customer history, order systems, inventory visibility and commerce flows.

Pandora’s broader digital transformation made this possible. Before introducing AI agents, the business had already modernized commerce, improved order management and strengthened its omnichannel foundation. Publicis Sapient helped Pandora build a more unified retail model across digital commerce, fulfillment and customer service, while also removing silos between marketing, commercial operations and technology. That foundation matters because the best AI experiences depend on connected systems underneath them.

The business outcomes show why this approach is resonating. Pandora reported 60% autonomous case deflection, allowing service specialists to focus more attention on higher-value interactions. It also reported a 10% uplift in Net Promoter Score driven by agent-first service. And with 22% of total sales handled online through Commerce Cloud, the digital channel is clearly a meaningful part of the commercial engine, not just a support layer.

For retail leaders, one of the most practical lessons from this work is that not every AI agent should do the same job. Service agents and shopping agents solve different problems and should be deployed at different points in the journey.

A service agent is the right fit when customers need speed, reassurance and operational clarity. Think order status, delivery updates, returns, refunds, care instructions and other repeatable post-purchase or support scenarios. In these cases, the biggest value often comes from reducing friction, lowering case volumes and creating better always-on support.

A shopping agent is the right fit when customers need help deciding. This is especially powerful in luxury and specialty categories where purchases are emotional, occasion-led or highly considered. Here, the agent should guide discovery, refine choices and surface relevant recommendations without making the experience feel mechanical.

In practice, retailers should ask four questions when deciding where to deploy each kind of agent:

  1. First, what is the customer trying to achieve in that moment: resolve, decide or explore?
  2. Second, what data does the agent need in order to be genuinely helpful: service history, order data, product attributes, behavioral signals or journey context?
  3. Third, what happens if the agent gets it wrong: minor inconvenience, lost sale or damaged trust?
  4. Fourth, where should human specialists stay in the loop: complex service recovery, high-emotion moments or premium consultations?
The best answer is rarely one universal agent. It is usually a portfolio approach, where different agents are designed around distinct customer intents and connected into a broader experience ecosystem.

That is where luxury and specialty retailers can learn from Pandora. Bringing the warmth of the store into digital channels is not about layering AI on top of disconnected experiences. It is about combining strategy, experience design, commerce, service and data into interactions that feel coherent and useful.

When done well, AI agents can do more than reduce cost to serve. They can extend brand voice, strengthen customer trust and create more guided paths to purchase. For retailers competing on service, storytelling and emotional connection, that is the real opportunity: not colder automation, but warmer digital experiences at scale.