"Under 180. Nothing black. With sleeves." She never touched a filter.

Conversational product discovery is when a shopper finds what they want by describing it in their own words, instead of clicking through filters, categories, and sort menus. The shopper says what they need — a budget, a colour they don't want, a feature they do — and the assistant narrows a real, live catalogue to match, holding every constraint as the conversation continues. Maya does this by voice or text: "a wedding dress under 180, nothing black, with sleeves" gets a matching shortlist from your actual products, and "now show me shoes for that one" keeps everything already agreed and adds the new request on top.

Last updated 2026-07-16

Where filters and search bars stop

A filter panel only understands the categories someone already built into it — price range, size, colour swatches. It has no field for "nothing black" as a preference layered on top of three other constraints, and no memory: pick a new filter and the old ones you cared about often reset with it. A search bar is worse — it wants a product name or a SKU, not a sentence. So the shopper with a real, specific need either wrestles the interface into something close enough, or gives up and leaves without ever seeing the product that was actually right for them.

Maya begins where the filter panel runs out of checkboxes. A filter is a print catalogue you flip through page by page, hoping the right item is somewhere in there; Maya is the assistant who actually listens, narrows it down with you, and tells you why something fits. The shopper doesn't learn your category tree — they just say what they want, in whatever order it occurs to them, and Maya keeps every constraint straight against your real, live catalogue.

What conversational discovery looks like on your store

It holds every constraint as the conversation builds — nothing gets dropped — "under 180 · nothing black · with sleeves" narrows the shortlist, and when the shopper adds "now show me shoes for that one," the dress constraints don't vanish — Maya keeps the first decision in view while helping with the second. A filter panel resets; the conversation doesn't.

It answers from your live catalogue — real stock, real price, never a guess — every match Maya shows comes from your actual product data at that moment, not a script that goes stale. So the shopper never gets shown something that's actually sold out or priced differently than what's on the page.

It explains WHY an option fits — it doesn't just return a list — "this one fits because it's under your budget and has the sleeves you asked for" — grounded in your real product data, not a generic recommendation. That reason is what turns a shortlist into a decision, the way a good assistant on the floor would talk you through it.

It greets the shopper by voice or text — in their own language, from the first word — the shopper can talk to it or type to it, and Maya opens in their language, not yours. On mobile, saying what you want out loud is faster than typing it into a search bar one word at a time.

It drives the shopper to the product, hands-free — once Maya narrows it down, it scrolls and navigates to the exact product page for them — they barely touch a thing, right up until they choose to add to cart or pay.

It follows the shopper's language, even mid-conversation — if a shopper slips from English into Italian halfway through describing what they want, Maya follows without asking them to start over — the discovery doesn't reset just because the language did.

It handles the hesitation and offers the right alternative — "it's a bit over budget" or "do you have it in another colour" gets an informed answer and a real next option, the way a floor assistant would — not a dead end. (Booking plan and up for objection-handling.)

It captures the shopper who leaves without deciding — the shopper who narrowed it down to two but wasn't ready to buy leaves a name and contact instead of vanishing — a recoverable conversation, with the exact constraints attached, instead of an anonymous bounce.

What happens when Maya doesn't know the answer?

The obvious worry: what if the assistant recommends something that isn't actually right — wrong size available, a feature it doesn't have? Maya checks every match against your real, connected catalogue before it recommends anything. When a shopper asks about something Maya can't verify — a spec you never listed, a combination you don't stock — it says so instead of guessing, offers the closest real option, and flags the question for you. You mark it handled, or teach Maya the answer in one sentence.

What that means for you: A recommendation is only useful if it's true. Maya doesn't get rewarded for confidently describing a product feature you don't actually have — so the shortlist a shopper narrows down to is always something you can actually sell them, not a return and a bad review waiting to happen.