Tell the store what you need. Get three picks and why.
A shopper types “espresso at home, tiny counter, under $700” and gets a short list from the store’s own catalog, each pick with the reason it fits — not a page of keyword matches. The bar that asks the question changes with the audience; the picks never leave the catalog. Switch the visitor below to see four shoppers.
Built with AI Product Finder — see the feature →
Gear for the kitchen you actually have
Espresso machines, grinders, cast iron and knives — chosen for small counters and daily use.
Vela Compact Espresso · $449Eight inches wide, so it fits beside a kettle on a short counter.Heats in 25 secondsSteam wandBest sellerPick 1Corsa Hand Grinder · $89Grinds fine enough for espresso, then goes back in a drawer.Ceramic burrsNo cordNo counter spacePick 2
Duo Grind & Brew · $679One machine instead of two, if the grinder is what will not fit.Built-in grinderSteam wandTrendingPick 3
A shortlist from your own catalog, each pick with its reason, behind a bar that speaks to each audience.
Feed it the catalog
The product feed you already import — names, photos, prices, stock — is what the finder picks from. It can also lean on the rankings your recommendation engine keeps, such as best sellers and trending.
Give each audience its own bar
The finder is an action in a campaign, so it targets like one. A gift version for December, a cast-iron version for the ad, a “compare” version for shoppers who keep coming back to the machines.
Let it pick, and say why
The shopper describes the need. The results sheet opens over the page with a short summary and a card per pick, each with one sentence on why it fits — and a box to ask a follow-up.
How the AI Product Finder works on this store.
The shopper describes the need, not a product name
“Espresso at home, tiny counter, under $700” names no product. The finder reads the constraint (a small counter) and the budget, searches the catalog, and comes back with three picks and a short summary of the trade-off between them.
Every card is a catalog row, and every reason is about the question
Name, photo and price come from the product feed, so a pick cannot carry an invented price or a product the store does not sell. What the AI adds is one sentence on why each pick fits what was asked, and up to three short points about it.
Rankings come from the recommendation engine
Labels such as Best seller, Trending and Most viewed are the engine’s own rankings, and those rankings only ever return items in stock. The “goes well with” row under the picks is where bought-together items appear.
The bar’s words follow the audience
The default bar, the gift version (switched on by the campaign’s December dates), the cast-iron version (switched on by the ad’s UTM) and the compare version (switched on by machine views in two earlier visits) are variations of one finder action. Each has its own title, invitation and chips; the question is always the shopper’s.
Other campaign content appears only when it helps
The reviews widget and a video-demo booking are attached to the results page as sections the AI shows when they are relevant. The shopper who asked what owners say gets the reviews; nobody asked for a person, so the booking stays hidden.
It does not talk about the shopper
The finder does not know names or orders, and it does not tell a returning shopper what they browsed. Past visits decide which bar they see; the picks answer only the question they typed.
See how it is built, and where else it fits
Questions about this example
Where do the AI Product Finder's picks come from?
From the store's own product feed. Each card shows the catalog row as it is stored: name, photo, price and link. The AI chooses which rows fit and writes the summary and the reason for each pick, so it cannot show a product or a price the catalog does not have. It can also use the recommendation engine's rankings, such as best sellers and trending, which only include items in stock.
How does the finder change for different visitors?
The finder is an action inside a Personyze campaign, so it is targeted like any other action. Each audience can get its own version of the bar, with a different title, invitation and example chips. In this example the campaign's dates, an ad's UTM parameters and category views from earlier visits decide which version a shopper sees.
Does the AI Product Finder know who the shopper is?
No. It answers the question the shopper typed from the store's catalog and pages, and it does not know names or orders. It does not tell a returning shopper what they browsed. It runs on the AI Chat Agent add-on, and it answers in the shopper's language.
Put a finder on your catalog.
Let shoppers say what they need in their own words, and answer with a shortlist from your own feed — each pick with its reason, behind a bar written for each audience.