Macy's
conversational genAI shopping assistant (multimodal agent with virtual try-on)
In March 2026, Macy's launched Ask Macy's, a multimodal conversational shopping agent based on Google Gemini Enterprise, deployed to 100% of its web and app visitors over a catalog of 2.5 million SKUs; in beta, revenue per visit was 4.75 times higher among the agent's users.
Key points
- Multimodal conversational shopping agent with virtual try-on on macys.com and the app.
- Google Gemini Enterprise for Customer Experience, catalog of 2.5 million SKUs.
- Revenue per visit 4.75x in beta, thousands of customers per day two months after launch.
- Evidence B, confirmed status: deployed to 100% of visitors, A/B measurement cited by Macy's.
Objective
Turn product search into guided discovery to grow revenue per visit and help the customer find what they are really looking for, from the hesitant browser to the buyer.
The deployment
Ask Macy's is a conversational shopping agent deployed on macys.com (desktop and mobile) and the Macy's app (iOS and Android). The customer describes in natural language what they are looking for - budget, occasion, color, style, size - and the agent responds like a salesperson: it suggests brands and trends, recommends products, completes the outfit (complete the look option for accessories). It is a multimodal agent that handles text and image: the customer can upload a photo for a virtual try-on that shows the garment worn, with different backgrounds, including in store. The system operates on a catalog of more than 2.5 million SKUs. Built on Google Cloud's Gemini Enterprise for Customer Experience after abandoning a first approach run over about six months, the agent was stood up in less than six weeks: first meeting on February 9, 2026, beta four weeks later with a small share of visitors and thousands of employees, moving to 50% of visitors the next day, then 100% a week later. The public launch took place around March 23-24, 2026, announced at the Shoptalk conference in Las Vegas, after an internal dark launch in December 2025.
Results Proof B
Revenue-per-visit figure (4.75x) from an A/B test cited by Macy's, picked up in Google Cloud's official customer story (vendor) and corroborated by two established press outlets (Fortune, Retail Dive). This is not a financial-results figure, hence level B rather than A.
How it works
Documented architectureThe stack in detail
- plateforme Gemini Enterprise for Customer Experience Google Cloud agentic platform, integrated with Macy's existing systems; multimodal agent (text and image) with virtual try-on.
How it runs, concretely
For ops teams-
1Capture intent in natural language AI
The customer describes what they are looking for; the agent asks framing questions (occasion, budget, size, style) rather than returning a list of results.
-
2Query the catalog and recommend AI
The agent crosses the intent with the catalog of more than 2.5 million SKUs to suggest products, brands and complete outfits (complete the look).
-
3Offer the virtual try-on AI
On photo upload, the agent shows the product worn, with different backgrounds, including an in-store variant.
-
4Supervise and adjust product / customer experience team
The team refines the tone and guardrails (for example adjustments to account for the weather, a friendlier tone) and monitors answer quality; a warning indicates the AI can be wrong.
The context provided by the customer in the conversation (budget, occasion, color, style, size) crossed with the real-time product catalog. If the catalog feed or availability data is missing, the agent recommends wrong or out-of-stock SKUs and loses its credibility.
How your customers perceive this type of use
Sourced studiesLes consommateurs n'acceptent pas les chatbots par defaut : 64% prefereraient que les entreprises n'utilisent pas d'IA dans leur service client (Gartner, 2024) et pres d'un utilisateur sur cinq du service client par IA n'en retire aucun benefice (Qualtrics, 2025). L'acceptation se construit sur trois conditions mesurees par Salesforce : savoir qu'on parle a une IA, pouvoir escalader vers un humain, comprendre la logique de l'agent.
Acceptance conditions
- Etre informe qu'on parle a une IA et non a un humain (pres de 75% le demandent, Salesforce 2024)
- Un chemin d'escalade clair vers un agent humain (45% plus enclins a utiliser l'agent IA, Salesforce 2024)
- Une logique de l'agent clairement expliquee (44% plus enclins, Salesforce 2024)
Red lines
- Rendre l'humain injoignable : c'est la premiere inquietude des consommateurs sur l'IA dans le service client (Gartner 2024) et 50% craignent que l'IA les coupe du contact humain (Qualtrics 2025)
- Remplacer le service client par l'IA sans alternative : 53% envisageraient de partir chez un concurrent (Gartner 2024)
Sources: Salesforce 2024 · Gartner 2024 · Qualtrics 2025
How to replicate
Inference, not sourcedData prerequisites
- Structured product catalog, up to date and queryable in real time (attributes, visuals, availability, price)
- Rich attribute reference set (occasion, style, color, size) so the agent reasons beyond the keyword
- For virtual try-on: the ability to process uploaded images within a clear consent framework
Org prerequisites
- A digital product team able to run an agent in production and iterate on tone and guardrails
- A platform partnership (here Google Cloud) or an internal AI team for integration and evaluation
- An A/B measurement setup to isolate the agent's effect on revenue per visit
- Photo data governance (biometric in the EU) and compliant AI transparency
Possible stack
- Google Gemini Enterprise for Customer Experience
- Or a conversational agent built on a multimodal LLM connected to the catalog via RAG
- Image-based virtual try-on component
- Native or third-party A/B testing tooling
First step: Verify that the product catalog is queryable in real time with attributes rich enough for an agent to reason by occasion and style, not just by keyword: that is what separates a real assistant from a search engine in disguise.
The same pattern, proven in other industries
Booking.com
genAI trip planning assistant
Burberry
genAI copilot for consultants (search, image recognition, complete-the-look)
Homes.com
natural-language voice and text search, then conversational assistant
Sources
- S1 A New Era for Online Shopping: How Macy's Built the 'Ask Macy's' AI Agent in 4 Weeks With Gemini Enterprise for Customer Experience Interested party archive pending
- S2 Macy's just launched an AI-powered shopping assistant. Customers who use it spend nearly 400% more Established press archive pending
- S3 Macy's introduces AI-powered shopping assistant Established press archive pending
An error, newer info, a source?
This page lives on its accuracy. If a figure has moved, if the deployment has changed, or if you have a higher-quality source, tell us. Every sourced correction is verified before publication.