Start free
Shopify · Shopping assistant

A good assistant
removes options.

Most store AI adds more things to look at. A shop assistant worth having does the opposite: asks the two questions that eliminate ninety per cent of the catalog, then puts one product in front of the shopper with a reason attached.

Install on ShopifyHow it differs from a chatbot ⟶
A forage route: one bee tracing a deliberate path between three blooms, rather than wandering
The problem

Your catalog is not too small. It is too wide.

A shopper who lands on a 400-product store and knows only roughly what they want has to do the merchandising themselves — open filters, guess at categories, compare across tabs. Most do not. They leave, and the store records it as a bounce rather than as a question nobody answered.

Search needs the word

Store search matches strings. A shopper who does not know your product vocabulary types the wrong noun and gets nothing, then concludes you do not stock it.

Filters need expertise

Facets assume the shopper already knows which attribute matters. Someone buying a gift does not know whether they care about weight, finish or drop.

Grids need patience

A grid shows everything equally. That is fair, and useless, when the shopper needs one answer rather than a fair sample.

The conversation

Ask, eliminate, resolve

Three moves, in order. The assistant is scoped to pre-purchase decisions — the one place it can be reliably correct — and hands everything else to a human.

Step
01

Ask what it is for

Not "how can I help?" — a question with discriminating power. Who is it for, what is the occasion, what is the ceiling. Two answers usually eliminate most of a catalog.

Step
02

Eliminate out loud

The assistant says what it ruled out and why. A shopper who sees the reasoning trusts the recommendation; one handed a product with no rationale goes back to the grid to check.

Step
03

Resolve to one variant

A specific size, colour and price that is actually in stock — then a real Shopify checkout link. No parallel cart, no re-picking options the shopper already settled.

What it is built on

Answers bounded by your catalog

Your products, kept current

Variants, prices, stock and your own copy ingest from the store and stay in sync, so the assistant never recommends something a shopper cannot buy.

Refusal over invention

Outside the catalog, it declines and offers a human. An assistant that guesses at a delivery date generates a refund and a bad review, not a sale.

Your checkout, untouched

It returns a Shopify checkout link. Your payment methods, your tax rules, your post-purchase flow — we never handle card details.

Handoff with context

Order, returns and account questions escalate with the conversation attached, so the customer does not start again from the top.

1
variant resolved, not a shelf of twenty
2–3
questions to narrow a wide catalog
Shopify
checkout — your rules, your payment methods
Related

The rest of the cluster

Store as agent tools

When the shopper is talking to an agent somewhere else, your catalog still needs to be callable.

Conversational commerce

Why a conversation converts differently from a grid, and what it costs to run one badly.

Agent vs chatbot

Scripted flows and a store agent are not the same product. The difference shows up in the questions nobody scripted.

Questions

Shopping assistants, answered

01What does an AI shopping assistant actually do?+
It narrows. A shopper arrives knowing roughly what they want and facing a few hundred products; the assistant asks the two or three questions that eliminate most of them, then puts one specific variant in front of the shopper with a reason attached. The measure of a good assistant is how few options survive the conversation, not how many it suggests.
02How is this different from product recommendations?+
A recommender ranks your catalog against a profile and shows a shelf. An assistant asks what the shopper is actually solving for and removes what does not fit. Recommendations are a merchandising surface; an assistant is a conversation that ends in a decision. Most stores benefit from both, doing different jobs.
03Will it make things up about my products?+
It answers from your ingested catalog — variants, prices, availability and your own descriptions. When the catalog cannot support an answer, it says so and offers a human rather than improvising. An assistant that invents a fabric composition or a delivery date costs more in returns and trust than it earns in conversions.
04Does it work for a shopper who does not know what they want?+
That is the case it is for. A shopper with a SKU in mind uses search. The assistant earns its place with the vaguer arrival — "something for a housewarming, under £40" — where the store has the answer but no interface that gets to it in under a dozen clicks.
05What happens at the end of the conversation?+
A real Shopify checkout link for the specific variant chosen, so the shopper lands in your normal checkout with your payment methods and tax rules. The assistant does not run a parallel cart or handle payment — it removes the distance between deciding and buying.
06Can it handle order and returns questions too?+
Those escalate. Order-specific and account-specific questions hand off to a human with the conversation attached, so the customer never repeats themselves. The assistant is scoped to pre-purchase decisions, which is where it can be reliably correct.

Stop making shoppers do the merchandising.

Install the agent, connect your catalog, and let a conversation do what a filter rail cannot.

Install on ShopifySee the agent tools