Upsell & Cross-sell

Next-best-product recommendation block (retail)

A shopper asks which jacket is warmer for hiking; the reply answers the question and shows the base layer most buyers pair with it.

Next best thingBehavioral triggerChatIn-appE-commerce & Retail

What it does

Retail shoppers reveal intent the moment they compare: "which of these is warmer for hiking?", "does this go with the boots I bought?", "what do I need to set this up?". A search box cannot read that intent and a plain text bot answers and walks away from the basket. This agent combines the live question with purchase history and renders a recommendation block inside its reply — the item that answers the question plus the companion most buyers add, each with a one-line reason grounded in fit, occasion, or compatibility. It is merchandising at the point of decision, native to the conversation, with every outcome written back to memory so the reasoning sharpens.

How it works

1
Trigger. mid-conversation, the shopper asks a need-revealing retail question — a comparison, a compatibility check, a use-case ("warm enough for winter hiking").
2
Decision. the agent reasons over the catalog and the shopper's purchase history — matching on attributes, occasion, and what pairs with items they already own — and picks the answer plus one or two genuine companions, each with a reason; if nothing pairs well, it answers and recommends nothing.
3
Action. the reply renders an inline recommendation block — product card, price, the reason — using the storefront's existing carousel and info-card components so it reads as part of the experience, in the chat widget or in-app.
4
Follow-through. taps, adds-to-cart, and dismissals are written back to memory; a dismissed item is not pushed again and an accepted one informs the next visit.

Example

CustomerWhich of these two jackets is warmer for hiking in winter?

AgentThe second one — it's rated to a lower temperature and blocks wind better, which matters more on exposed trails. Most people pair it with a merino base layer for the cold mornings, and that base layer also goes with the trail trousers you bought last month. Want either added?

CustomerAdd the base layer.

AgentDone — it's in your basket in your size. The jacket's there too if you want to compare them side by side before you decide.

Configuration

How the agent is wired for this use case.

Triggeran in-session classifier flags a need_revealing_message (comparison, compatibility, or use-case question) in the chat or in-app widget.

Tools & actions

Product catalog service · read attributes, stock, and compatibility/occasion metadata; select the answer plus genuine companions
Order / purchase history · read prior purchases to ground "pairs with what you own" reasoning
Cart API · add the accepted item, in the right size or variant, to the shopper's basket
Customer memory · read past dismissals and preferences; write back taps, adds, and dismissals
Messaging channel · render the recommendation block via the storefront's carousel and info-card components

Autonomyrecommendation selection and the inline block run unattended under judge gating on the recommendation turn; adding to cart is a one-tap shopper confirmation. No money moves — checkout stays in the storefront's own secure flow.

Channelschat · in-app

Escalationan out-of-policy request, a stock or fulfilment exception, or a question the catalog cannot answer hands off to a human with the conversation context.

What you need

The inputs this use case runs on. Your channels stay yours; the agent supplies the judgment.

Signals

need-revealing retail message classified in-conversation — comparison, compatibility, or occasion intent

Data

product catalog with attributes, compatibility and occasion metadata, and stock; purchase history; customer memory of preferences and past dismissals

Guardrails

recommend only on genuine fit — silence beats a forced pairing; dismissal memory prevents repeat pushes; judge gating on the recommendation turn; checkout stays in the storefront's secure flow and the merchant keeps owning the surface

Metrics it moves

attach-rateup, as the matched companion rides inside the answer the shopper asked for
conversion-rateup on recommendations versus interruptive banners, because the context earns the offer
aovup when the block surfaces the compatible add-on alongside the comparison