Every storefront agent is trained on your brand specifically — this article explains how that happens, from the moment your agent is set up through to how it keeps learning.
Learning who you are
Before your agent ever talks to a customer, it goes through a research process built entirely around your brand.
It starts by reading everything publicly available about you — your website, your blog, your social presence — to build up a picture of your brand's story, tone of voice, hero products, and policies like returns and shipping.
Then it goes further than a website alone ever could. You'll be asked to fill in a short brand form covering things the agent can't pick up from web scraping: how you want the agent to handle specific situations, phrases you'd never want it to use (and ones you'd love it to), the personality behind your brand, and sizing detail specific to how your products actually fit. Anything you tell us here takes priority over what the agent found on its own.
💡 The agent's tone is shaped by this too — so it doesn't just know your products, it sounds like your brand.
Knowledge Packs
All of this is organised into topic-based knowledge categories called Knowledge Packs. The agent creates a Knowledge Pack on your sizing, returns, shipping, brand story, and more — so that whatever a customer asks, the agent draws on the right piece of brand knowledge to answer it, rather than giving a generic response.
Example: One brand shared transcripts from their regular customer styling calls, which became a bespoke pack of styling advice that only their agent has access to — turning real conversations their team already has into something the agent can offer at scale.
Built to stay accurate
Your agent is built to answer only from what it actually knows about your brand. It won't invent bestsellers, make up shipping promises, or claim things it can't back up — every answer is checked against real brand content before it's trusted.
Getting to know your catalogue
Your agent also builds a deep understanding of your product catalogue itself. Every product is read and organised into a consistent set of categories, so the agent always knows what it's looking at and can compare like with like across your entire range — even if your own product data is a little inconsistent day to day.
This same understanding is also what allows us to automatically manage more advanced storefront features for specific product types, for example, turning off Virtual Try-On for categories where it isn't appropriate.
An agent that keeps improving
Your agent's setup doesn't stop at launch. A member of the Swap team reviews and fine-tunes everything before it goes live, and once your full catalogue is up and running, we revisit and sharpen the agent's responses against your real, live inventory — so it only gets more accurate over time.
Curious to see it in action? Book a demo to learn more about Agentic Storefront, or if you're already using it, just ask your storefront agent a question about your own products — that's the best way to see what it already knows.