Your assistant can now offer tappable quick replies after it answers

Shopping assistant agents can now offer quick replies: up to three short, tappable suggestions that appear right above the message box once the assistant has delivered something: a product carousel, a comparison, a product detail, or a factual answer, including a "not in stock" or "no such code" kind of answer. Tapping one sends its full message on the shopper's behalf, exactly as if they'd typed it. No more retyping "what about a cheaper one?" after every answer.

A zero-result search gets the same treatment: instead of a dead end, the assistant can offer a wider search, a different budget, or a related idea.

Suggestions are also checked before the shopper sees them. Some are refused because your store can't support them in that moment: "go to checkout" with an empty cart, "compare these" with a single product on screen. Others are refused because the answer isn't there: each suggestion says where its answer would come from, and when that's one product's catalog fields, its description, or what your shop published in its policies and FAQ, the assistant reads that source first and drops the suggestion if the promised fact isn't in it. "Your delivery times" doesn't get offered when your shipping page lists carriers and options but never states a timeframe.

The assistant drafts more suggestions than the row can hold, so a refused one is usually replaced by the next best rather than leaving a gap. The more specific your product descriptions and published policies are, the more of them survive. The check has limits, though: it doesn't apply to a chip that adds to the cart, opens the cart, or runs another catalog search, and when the source can't be read (it doesn't come back in time, or the product behind the chip can't be resolved), the chip is shown anyway. So it makes a suggestion the assistant then can't answer rarer, not impossible.

Every chip renders identically (plain text, no icon, no highlighted "best pick"), so nothing nudges the shopper toward one option. Chip text is set at the same size as the conversation's own messages. The row has no close button; typing a new message (or the assistant moving on) is what dismisses it, and it never appears while the assistant is still working or while a clarifying question is pending.

On a wide chat panel the chips wrap onto a second line when they don't all fit; on a narrow one they stay on a single sideways-scrolling line, so the row costs one line of height instead of three and the conversation stays visible.

New agent versions have quick replies enabled by default, alongside a new closing style: instead of ending on a prose question that lists its own branches, the assistant now states its own judgment and lets the chips carry the next-step choice. An existing agent version is unaffected: the new behavior only applies to a version created from now on. You can turn quick replies off per version like any other chat tool.

Quick replies stay silent right after a confirmed add-to-cart except for a "go to checkout" or a delivery/returns reassurance, never a chip that pulls the shopper back into browsing right when they've decided.