Cloud Atelier: How Gemini Enterprise is helping restyle the retail playbook

We’ve all been there: staring blankly into an open suitcase, asking the question, what do I pack?

This year at Cannes, we are thrilled to showcase Cloud Atelier — a destination-based, virtual shopping experience that highlights how retail brands can turn this classic dilemma into an exciting moment of product discovery.

Built on Gemini Enterprise Agent Platform, this industry-focused demo invites users to virtually try on a travel wardrobe tailored to their trip details, style preferences, and up-to-date data like weather conditions and local trends.

Distinct from Google's Try on you feature — which appears in eligible shopping results on Search — Cloud Atelier brings together the combined power of agentic commerce, personalized creative, and real-time execution to create a context-aware, shoppable lookbook.

In doing so, Cloud Atelier demonstrates how retail brands can leverage Google Cloud’s AI stack to move beyond traditional product recommendation engines and deliver more directly on customer intent — turning the what to pack question into a seamless path to purchase.

Agents transform natural language into next-level looks

Bringing a personalized virtual dressing room to life requires a lot more than a single prompt. Let’s break down exactly how Cloud Atelier utilizes an agentic team — created with Google ADK and deployed to Cloud Run — to guide a user from sharing their travel plans to a custom wardrobe recommendation.

In the first step of the demo, the user describes an upcoming trip in their own words. Behind the scenes, Gemini’s native capabilities process the natural language response, extract key data (where and when they are traveling), and determine if any follow-up questions are necessary to fill in missing gaps.

While the user completes the onboarding process, the orchestrator of the agent team — the Wardrobe planner agent — deploys various specialized research sub-agents to complete tasks in the background. Depending on task complexity, the agents rely on Gemini Flash or Gemini Flash-Lite models and are equipped with specialized features such as the Google Search tool, which ensures that all results are grounded in real-world accuracy.

• Weather research agent: Accesses the Weather API tool to pull accurate forecast data for the user's travel dates.
• Travel research agent: Proposes culturally authentic destination activities (e.g., “ride a Vespa” or “get gelato” for a sunny weekend in Rome, Italy).
• Trends research agent: Researches local fashion trends by searching external web sources.
• Location research agent: Retrieves hotel data and location images via the Google Maps Places API.

Meanwhile, the user shares their style preferences by selecting from a curated library of aesthetic categories (such as outdoorsy, business casual, or preppy). To finish, the user snaps a full-body photo or, alternatively, chooses from a selection of avatars.

Once the research sub-agents return their results, a new team of sub-agents goes to work while the user reviews travel tips and an itinerary personalized for their trip. First, the wardrobe planner agent uses its reasoning capabilities to synthesize the research data and generate relevant product recommendations (e.g., a packable rain shell for a trip to Scotland in the fall).

Then, a Wardrobe search sub-agent takes over to cross reference those recommendations against the product database to match them with real, inventoried items.

To do this, the sub-agent uses Vector Search, which bridges the creative flexibility of Gemini reasoning with the accuracy of enterprise-grade semantic search. Instead of just looking for exact keyword matches, Vector Search compares items across thousands of abstract traits — so it can tell that a trench coat and a rain shell are closely related. For example, if the sub-agent is searching for "waterproof boots," the system can surface "weather-resistant footwear" or find the best substitute if that item is out of stock.

After finding the best matches, the sub-agent pulls the relevant item data (like product name, price, and description) from BigQuery and item images from a Google Cloud Storage bucket.

Those product images and the user’s photo are passed along to the Content generation sub-agent. Leveraging Nano Banana 2, the sub-agent visualizes the user wearing those looks against backdrops of their travel destination — delivering flattering styling that’s true to their body type.

Then comes the best part: the interactive digital lookbook. The user can explore five customized, shoppable outfits complete with interactive product cards. After adding their favorite look to their cart, the user can scan an on-screen QR code for a digital takeaway that includes a video of their favorite look in motion (generated by Veo 3.1) and other helpful details like a product checklist and hotel recommendations.

By weaving all these digital touchpoints together, Cloud Atelier highlights how a virtual try-on experience can go far beyond simple wardrobe suggestions — offering retailers an adaptable foundation to turn customer inspiration into a ready-to-buy reality.

Tomorrow’s tech is setting up shop today

A number of retail brands are already utilizing Gemini Enterprise to drive real business outcomes:

• Kohl's (specialized gifting agent): Kohl's is combining Looker's semantic layer with Gemini capabilities to build conversational analytics. The retailer also launched a specialized Mother's Day gifting agent (using Gemini Enterprise for Customer Experience) that dynamically helps shoppers discover, select, and "add-to-cart" highly relevant gifts based on conversational prompts.
• Lowe’s (kitchen inspiration app): Powered by Gemini 2.5 Flash, the Kitchen Inspiration experience turns a single photo into a personalized design exploration. It generates kitchen styles tailored to your space and helps you define a clear design direction in minutes, so you arrive at your next conversation with a Lowe's designer knowing exactly what you want.
• Futuriza (virtual fitting rooms): This retail-tech startup utilizes Nano Banana and Veo to power plug-and-play AI virtual fitting rooms for apparel brands. For their retail customer Arm Fitness, in just 1 month, they generated over 20,000 virtual try-ons, with 38% of users adding items to their cart, and boosted the sales conversion rate from 3.5% to 4.7% (a 35% increase).

These examples show that the future of commerce isn't just about putting items on a digital shelf. It's about understanding exactly where your customer is standing, where they're headed next, and helping them step out in style.

And that future is yours to create. We’ve made a Cloud Atelier code starter kit publicly available — so you can explore how it works, adapt it to your own catalog, and start creating context-aware shopping experiences of your own. What will you build?