Training your own multimodal model from scratch takes a lot of time and money. You need data, compute, a team, testing, and then months of work before the result is good enough to show to an actual customer rather than just demo internally. Somewhere deep inside the research labs of very wealthy companies, people are probably doing exactly that.
A lot of brands are taking a different route. Ralph Lauren, for example, launched Ask Ralph.
Ask Ralph appeared in the Ralph Lauren app in the US in September 2025. Users can type where they’re going or what they want to wear, and the system suggests several outfits from Polo Ralph Lauren. You can open individual items and go straight to purchase. At launch, the service was limited to Polo Men’s and Women’s in the US.
Ask Ralph is built on ready-made Microsoft services, but a raw LLM has no idea what Ralph Lauren considers a good look. So the system was given the brand’s own material: product imagery, PDP images, editorial campaigns, style guides, and brand language. At launch, Ask Ralph was limited to Polo Men’s and Women’s, which gives you a pretty good idea of where the real work starts: every new line has to be prepared with the right products, images, descriptions, and styling context before the system can work with it properly.
In engineering terms, this looks like a familiar RAG pattern. Microsoft says Ask Ralph uses conversational AI and natural language processing to understand open-ended prompts and the context of a conversation. From there, the system needs to find actual SKUs, combine them into something that makes sense for the brand, and check that those items are still available before returning the final answer.
The heaviest tech comes from the outside. Ralph Lauren didn’t have to build its own model from scratch. It had to make an existing one work with its own assortment.
If this approach becomes cheap and simple enough, almost every large brand could end up with its own agent.
But will anyone actually use them?
We’ve seen a similar cycle before with live chat. Once it became easy to add chat to a website, it appeared almost everywhere. At first it looked like a useful new feature. Then it became another button that many people close automatically.
Maybe branded AI agents really will become a new way to shop for clothes. Or maybe, a few years from now, every brand will have its own stylist, consultant, and shopping assistant, and we’ll learn to close all of them just as quickly as we close “Chat with us” today.
