"Our taxonomy is developed by our master stylist, our master perfumers, but the way a fragrance expert speaks is not necessarily the way a typical consumer speaks," says Curro. "Instead, we have to figure out how to bridge that gap, and that's where a broader agentic system comes into play, powered by Google Gemini." 

The core of the system is an interactive AI agent that uses Gemini's reasoning and multimodal capabilities to evoke necessary points of inquiry by asking about a customer's memories, sensory experiences, and the identity they want to express through scent. The agent then translates this subjective language (that is, "a scent that reminds me of a vintage library") into a structured query format that is fed into a ranking engine, built on Agent Platform, to select the best fragrance candidates. Finally, the agent crafts a personalized response explaining why these scents are a match. 

Google Cloud forms the backbone of the AI Scent Advisor: Cloud Run services and worker pools, along with Pub/Sub, enable continuous data processing at scale. Firestore and BigQuery serve as a powerful operational and analytical data foundation, with Agent Platform and Gemini providing the core AI logic and conversational flow. Gemini's controlled generation also made it possible to define a precise format and structure for AI responses, ensuring reliable, readily usable outputs that could be easily integrated into the AI Scent Advisor's complex reasoning workflows. 

Crucially, Google Security, including Model Armor, has enabled Jo Malone London to implement and enforce safety policies that can handle nuanced terminology like "nude" for discussing skin tone or foundation that might be problematic in non-beauty contexts. 

"Google Cloud knows how to make you feel like you're customer number one," says Curro. "No matter what we need on whatever timeline it may be, our Google Cloud team has been there making sure that we get everything we need to deliver the business value that is critical to any company trying to lead the charge in technology, in AI." 

"To do systematic analysis, distill out the key techniques, we turn to Google Gemini to make that a scalable process," says Curro. "We were able to extract the key vernacular, really hone in on the key factors that make the Jo Malone London stylist 'tell, explain, describe' framework work, and bring that into the Jo Malone London AI Scent Advisor."

With Gemini, the team was able to then codify and simulate a voice that captures the same spirit of curiosity and creativity of real stylists while providing expert-level, personalized guidance. 

"The metrics have been fantastic. We've been driving conversion. We know customers are really, really satisfied with it," says Curro. In addition, Jo Malone London's research has shown that customers are going into physical stores to discuss and try the recommendations they received, demonstrating the potential of digital investment to enhance the overall brand experience. 

The Jo Malone London AI Scent Advisor is not only setting a new bar for online fragrance discovery, it's also fueling the brand's future innovation pipeline. Using proprietary natural language analytics, built on Google Cloud, Jo Malone London is gaining deeper insights into consumers and their preferences to deliver products and experiences that help customers express their personal beauty—in a way that is uniquely their own.