ZenAI Shop Assistant

Your best salesperson.
On every page.
At 3am.

ZenAI answers your shoppers from your live catalogue: real stock, real sizes, real prices, your own policies. It never invents a product you don't sell. And when the answer is no, it says no.

ZShop Assistant Live
70/30 fridge freezer, frost free
Here are some 70/30 fridge freezers with frost-free technology:
Hotpoint · €599.99Whirlpool · €899.99Indesit · €449.99
Would you prefer freestanding, or built into your kitchen?
Built in. Under €500
I'm afraid the built-in 70/30 models we have are above €500. The built-in options start at €899. Would you like to see freestanding 70/30 models under €500 instead?
Yes, show freestandingSee built-in anyway
REAL CONVERSATION · 11 AUG 2026✓ GROUNDED IN LIVE STOCK
1 in 20
online orders involved a shopper who used it
99.6%
of replies delivered clean
1.6s
average time to an answer
24 hrs
from catalogue to live on your storefront
The gap

The questions arrive when the shop is shut.

They have one question. Does it come in a 10 wide, is it actually in stock, will it fit through my door. And nobody is there to answer it. So they leave, and you never find out what they wanted. That isn't a traffic problem. It's an unanswered-question problem.

The 3am question

Browsing doesn't stop when your team clocks off. ZenAI answers on nights, Sundays and bank holidays, in the shopper's own language, without waking anyone up.

Always on

The size question

Shoppers don't know their size across brands, and a guess is a return. ZenAI translates between official brand charts so they order once.

Fewer returns

The "where's my order" question

Order-status queries were answered in chat instead of landing in your inbox. Your team gets its morning back.

Support deflected
Why it's trusted on the storefront

It says no.
That's the whole point.

Most store chatbots are a language model with your logo on it. They will happily invent a product, a size, a price or a returns policy, and every one of those is a refund, a complaint, or a customer you never see again.

ZenAI can only speak from what it can actually read: your live Shopify catalogue, your real stock levels, your published policies. If the thing doesn't exist, it tells the shopper, and offers the nearest thing that does.

In the conversation above, the shopper asked for a built-in model under €500. There wasn't one. ZenAI said so, and kept the sale alive by offering the freestanding range instead.
  • Grounded in live stock.Every product it names is read from your catalogue at the moment it answers, not from a monthly export.
  • Your policies, quoted.Shipping, returns and click-and-collect come from your own published pages. It never drafts terms you didn't write.
  • Guardrails per category.Pharmacy gets clinical suppression and a hand-off to the pharmacist. Electrical gets spec discipline. Footwear gets fit and width.
  • Scored, not assumed.An independent AI judge grades real conversations on helpfulness, relevance, accuracy and honesty. What it flags goes on our fix list.
  • You keep the keys.Your catalogue, your tone, your rules. Scope it to one page to start, and turn it off in one tap.
What it does for you

Six jobs your team can't do at 3am.

One assistant, tuned per store. No scripts to write and no decision trees to maintain. It reads your catalogue and gets on with it.

Finds the right product

Natural questions in, real products out. "Something for a knee replacement I can shower with." "Frost free, 70/30, built in."

Live catalogue search

Translates sizes across brands

ASICS 9.5 to New Balance 10. Built on official brand charts and gender-aware, so shoppers order once.

Brand-to-brand sizing

Runs a guided finder

A few taps and a big catalogue becomes three real options. Fit finder in footwear, supplement and shade finders in health and beauty, appliance finder in electrical.

Tuned to your category

Answers stock honestly

In stock, out of stock, last pair, wrong width. It checks before it claims, and it never oversells your shelf.

Stock-gated

Quotes your policies

Delivery times, returns windows, click-and-collect, free-delivery thresholds, all read from your own published pages, never invented.

Policy-grounded

Hands over cleanly

When it's a human job, like a prescription, a complaint or a bespoke quote, it stops and routes to your team with the context attached.

Knows its limits
Receipts, not vanity metrics

Six live stores. Sixty days. Real orders.

Four verticals, two currencies, one engine. Every figure below is read straight from production telemetry and the stores' own Shopify sales data. Nothing modelled, nothing rounded up.

1 in 20online orders across the three stores involved a shopper who had used the assistant
Why we count orders, not impressions

Most chat vendors quote an engagement rate against the number of times their widget loaded. We don't, because that denominator is worthless on a Shopify storefront: it counts bots, scrapers and preview traffic, and it varies by more than 2× between our own stores depending on where the widget sits.

Orders are the one thing a bot never does. So both halves of this ratio come from order records. ZenAI's attribution on one side, the store's own Shopify sales data on the other. It's the least flattering number we could have picked, and the only one that survives the question.

Running & footwear
United States · live since June
The heaviest user of sizing and fit. Shoppers who talked to it before checkout spent about 10% more per basket than the store average.
Pharmacy
Ireland · three stores, live June to July
Need-led rather than product-led. Clinical questions trigger a pharmacist referral before any product, and the infant-formula age gate holds under real pressure.
Electrical & sports
Ireland · live since July
Spec-heavy questions where a search box has nothing to say. The hero conversation on this page is a real one from an electrical retailer.
The honest read. Three of the six stores have a full 60 days of order attribution behind them; those are the numbers above. The other three went live in late July, so they have conversation data but not yet a complete attribution window. We have left them out rather than guess, and we will publish them at 60 days. Store names, revenue and per-store commercials are not published here because they are our customers' numbers, not ours. We will walk you through the full detail on a call.

Bigger baskets when they talk to it

At the running store, shoppers who had a real exchange with ZenAI before checkout spent around 10% more per order than the store's overall average.

Conservative window · chat-engaged only

1.6 seconds to an answer

Average reply time across every assistant turn in the window. Fast enough that shoppers keep typing instead of closing the tab.

Measured, not promised

99.6% of replies land clean

Counting both hard errors and any reply that came back empty. On a storefront the failure that matters isn't a slow answer, it's a broken one.

Production reliability
How it goes live

Reads your catalogue. Live in 24 hours.

No replatforming, no scripts, no chat trees. We prep your product data, point the assistant at it, and a human signs it off before a shopper ever sees it.

Step 1

We prep your catalogue

Product types, tags and Google categories cleaned up so the assistant can actually find what you sell. One store went from 4% to 99% correctly categorised.

Step 2

We build your store brain

Your tone, your policies, your vertical's guardrails, your delivery and collection rules. Written once, applied to every answer.

Step 3

We go live on your storefront

Installed and branded to your theme. Start scoped to a single page if you'd rather dip a toe. Off in one tap, always.

Step 4

We score every conversation

An independent judge grades real conversations every week. What it flags, we fix, and you see the fix list rather than a dashboard to babysit.

You stay in control of the whole thing.

We prep the catalogue, write the store brain and read the answers ourselves before it goes anywhere near your storefront. Nothing ships on a guess. Scope it to a single page to start, and switch it off in one tap whenever you want.

Give every shopper the person
they'd have met in your shop.

The same engine is already answering thousands of real shoppers across running, pharmacy, electrical and sports retail.

Official Shopify App · 5.0★ · Live in 24 hours
How these numbers are measured. Every figure on this page is read from ZenAI production telemetry and the stores' own Shopify sales data for 15 June to 13 August 2026, across the three stores with a full attribution window. 1 in 20 compares orders ZenAI attributed against total orders in Shopify for the same stores and the same dates; it is order-to-order, so bot and scraper traffic cannot reach either side of it. An attributed order is one where the assistant was present and interacted with in the seven days before checkout, which makes it an influence measure rather than a last-click one. The basket comparison uses the narrower chat-engaged definition, a genuine two-way exchange before checkout, against the same store's overall average order value in the same window. 99.6% counts both hard errors and empty replies as failures. 1.6 seconds is the turn-weighted mean reply time. Internal and test conversations are excluded throughout. We do not publish engagement rates measured against widget loads, because that denominator includes non-human traffic and is not comparable between stores.