Sierra says retailers should run one shopping agent that finds the product, closes the sale and handles the return.
The tell is a permission rule. A retailer might let a personal agent start a return but not change payment details, Sierra said. That small boundary is the whole pitch for Poppy, its open standard for personal agents and brand agents to work together.
Poppy was developed with 100+ businesses including Muse, Meta’s AI agent, Genesys, Shopify and Walmart, Sierra said. Consumers choose what their personal agent can access and retailers decide what it can do on their site.
That control matters because Sierra wants its own brand agent to be the front door. When a fit question comes up, the personal agent calls the brand agent for the answer.
Poppy puts Sierra Context Engine behind the chat
Don’t wait to be asked. That’s the operational shift.
Most shoppers still won’t open a chat window on their own, Sierra said. So the agent should start it with a tap-to-ask prompt like “Will it fit over a fleece” on a rain jacket page. The answer pulls product details, reviews and past purchases, then recommends the next size up, checks stock and adds it to the bag.
Sierra’s Context Engine connects catalog, commerce systems and customer records with what the shopper says, Sierra said. Save-the-sale lives there too. If the color is wrong, the agent can suggest a better one before it processes a refund.
Building that is pitched as no-code work. Ghostwriter takes SOPs, transcripts, whiteboard photos, audio recordings or a plain-English goal and builds a production-ready, multilingual, multichannel agent with guardrails, per Sierra’s site. Teams will soon track shopping performance directly in Agent Studio, Sierra said.
You’ve seen where this breaks. Product data is messy, size curves lie and reviews contradict the spec sheet. If Sierra can keep the permission tight and the context current, retailers keep more baskets.





