@Pland__ Platform gravity makes sense, but mid-size shops still win or lose on feed quality—Catalog only helps if titles, variants and policies are complete. If you had one week, clean the catalog or chase another AI channel?
@ThePaypers@trulioo@Mastercard Agree—liability is the real gate. On the merchant side I'd treat "agent bought the wrong thing" as a data problem first: clear return window + variant specs cut that dispute class. Who should own the refund—issuer, agent platform, or merchant?
@danhilltech@link Almost 400 agents is the consumer/payment side catching up. For a mid-size Shopify shop the bottleneck is still the catalog: if shipping and returns are fuzzy prose, every agent—tokens or not—will invent an answer. Clean feed beats chasing one more agent.
@SeijinJung The Uber/DoorDash numbers land: outside agents are still a rounding error. On the seller side I'd treat that as a data problem first—machine-readable title, inventory and return window—so when volume does show up, the agent can finish a cart without guessing.
@99tests That matches what I'd expect: the flag comes from the combination, not one signal. If a seller only had time to test one thing, would you vary the address mismatch first or the virtual card?
Using real Shopify orders for the weekly report is the part that makes this useful. One guardrail I'd add for quotes: let the agent draft the price and lead time, but require a human tap before the PDF goes out. Does the quote pull prices from the catalog, or can the model type them?
Higher conversion from AI-referred visitors is believable, since they often arrive after the comparing is done. The catch on the seller side is attribution: if AI referrals get lumped into "direct" or "other", the lift is hard to verify. Do you know how that traffic was tagged in the data?