Upselling and cross-selling with an AI sales assistant: what to suggest, and when

Upselling means suggesting a higher-priced or upgraded version of what the shopper is already looking at. Cross-selling means suggesting a different item to buy with it. In a chat, both work when the suggestion answers something the shopper has told you, and both fall flat when it's a stock line that appears whatever they said.

A website widget shows the same "you might also like" strip to everyone. A conversation knows this shopper is buying a gift or needs it to work with something they own, so the suggestion can fit.

A suggestion helps when it answers a question

The add-ons that land are usually the ones a shopper was about to ask about. Do I need a filter for this? Is there a case that fits? If the suggestion answers that question, the shopper hears advice.

Upselling works the same way. "The 2-person tent is fine for one, but if you camp with a dog the 3-person gives you the room" is useful. "Have you considered our premium range?" isn't, because it has no reason in it. And sometimes the honest answer is the cheaper option, which is worth saying when it's true.

Three moments to suggest something

When a shopper lands on a product page, one relevant add-on can sit beside the chat button. It's easy to ignore, and it tells them the shop knows what goes with what.

During the conversation, the best moment is right after you've answered their question. They've just told you what they're using it for, so the add-on can be specific.

After they add something to the cart, a short note about the refill or the matching part saves them a second order later.

How Agentmatica decides what to suggest

You set the rules in Training, under Pairings & upsells. There are three kinds. "When a shopper is looking at the espresso grinder, suggest the tamper" is a suggest rule. "Only suggest the travel case with the compact model" is an only-with rule. "Never suggest the decaf beans with the espresso kit" is a never-with rule. We draft a few pairings from your catalog, and you can edit any of them.

On a connected Shopify or WooCommerce store you can turn on Suggest a complementary product. On product pages the agent then offers one item, drawn from your pairing rules and your live catalog, as a small bubble beside the launcher if the chat is closed. It's off by default and doesn't use your reply quota. With no pairing rule that fits the page, it stays quiet.

In conversation, the agent follows your Sales notes, and your only-with and never-with rules still apply. Its sales coaching tells it to suggest one complementary item when it helps, and to keep the reply to a few short sentences. When the shopper says yes on a Shopify or WooCommerce store, it adds the item to the cart with the right variant and hands over a checkout link.

Across all proactive messages, the widget shows at most one card per page and two per visit, so a shopper browsing ten products doesn't get ten pitches.

What makes a suggestion feel like selling

The usual culprit is a suggestion with no reason attached. "Customers also bought" is a fact about other people, while "the 1 kg bag works out cheaper per cup if you drink two a day" is about this shopper. Pushing the priciest option every time does the same damage, and shoppers spot the pattern quickly.

Write your first five pairings

List your five best sellers and the one thing each one most often needs. Add them as suggest rules, then read the conversations where the agent offered them. If shoppers keep saying no to a pairing, change it.

Questions store owners ask