Shopify + AI support: a 5-minute setup guide

Going live on Shopify is three steps: create a workspace, upload your first documents, paste the widget into your theme. That is the five minutes. Testing it like a difficult customer takes as long as you give it - and you should give it an evening before the widget meets a real one.

This is the exact setup we walk new Shopify merchants through. No developer, no code - if you can paste a line of text, you can finish this before your chai goes cold. Set a timer.

Step 1: Create your workspace

Sign up, name your workspace after your store, and pick your brand colour - the widget preview updates live as you type. Set your greeting message to match how you'd welcome a customer in-store. Keep it short; "Hi! Ask me about sizing, shipping, or returns" outperforms a paragraph.

Step 2: Upload your knowledge

Drop in your five core documents: shipping policy, returns policy, size guide, FAQ, and store details. PDFs, Word files, or pasted text all work - everything is extracted, chunked, and embedded automatically. If your policies currently live only in your head, open a blank document and write the answers to your fifteen most-common questions. That single file is worth more than your entire product catalogue.

Step 3: Install the widget

Copy your one-line script tag. In Shopify admin, go to Online Store → Themes → Edit code, and paste it before the closing </body> tag in theme.liquid - or use the one-click plugin and skip even that. Save, reload your storefront, and the chat bubble appears bottom-right in your brand colour.

One script tag. Simpler than installing Google Analytics - and it talks back.

Then: test like a difficult customer

Now try to break it. Ask about shipping to a specific pincode, a return on a sale item, a size you don't stock. Check that every answer cites the right document. Then ask something your documents don't cover - you should see it escalate gracefully and ask for contact details rather than guess. That escalation lands in your email with the full transcript.

Finally, set your escalation email, invite a teammate if you have one, and you're live. Tonight, when the 10pm "do you ship to Indore?" message arrives, someone will answer it in four seconds - and it won't be you.

What to upload for a Shopify store

Five documents cover the overwhelming majority of pre-purchase questions an Indian store gets. They are the same five the e-commerce walkthrough recommends, and most stores already have them written.

Six questions to test with

Test the way customers actually type, not the way your FAQ is written. These six exercise the paths that matter.

1. A shipping question naming a real city you serve - and one you don't. The second should get your actual coverage policy, not an optimistic yes.
2. A return on day 10 of a 7-day window. The bot should say no, politely, citing the policy.
3. A COD order below your minimum, if you have one.
4. A sizing question your size chart actually answers.
5. Something your documents don't cover at all. The right answer is an honest refusal and an offer of a person - if it improvises instead, read this first.
6. A complaint - "my order arrived damaged." That must escalate with contact details, never get a policy quote. If any of the six misbehave, the fix is almost always a document edit, not a settings hunt.

COD, GST and the India-specific questions

Global Shopify advice skips the questions Indian stores actually get. Does COD work below Rs 999? Can I get a GST invoice against my company's GSTIN? Do you deliver to this pincode? These are decided pre-purchase questions - the customer who cannot get the COD answer buys elsewhere tonight.

Each has a home in the five documents above, but only if you state them explicitly. Write the COD minimum as a number. Say which pincode ranges or cities you serve rather than pan-India, which answers nothing. And spell out the GST invoice process, because B2B buyers ask before every order and it is the easiest sale-loser to fix in one paragraph.

A placement note while you are in the theme: the widget loads asynchronously and stays under 50KB, so leaving it on every page costs nothing in speed scores. The pages where it earns most are the product page - where the sizing and delivery questions happen - and the cart, where a doubt left unanswered becomes an abandonment statistic. If you run a separate policies page, leave the widget there too: a customer reading your returns policy has a returns question, and the widget answers it in their words instead of making them parse yours.

Sale season: the stress test you can schedule

Diwali, end-of-season, a creator post that lands - Shopify stores do not get steady traffic, they get spikes. The support load spikes with it, and this is where the setup you did in an evening earns its keep: the same fifteen questions arriving fifty times a night get answered without anyone staffing for the surge.

Two preparations matter. Upload the sale terms as their own document before the sale starts - dates, discount rules, what is excluded, whether COD applies to discounted items - and delete it the day the sale ends, so the bot never quotes an expired offer to a January customer. And if you are comparing tools, note what a spike does to a metered bill: the cost arithmetic is ugliest in exactly your best week. The festival playbook covers the fuller checklist.

What this does not replace

Honest scope: the widget answers questions from your documents. It does not process refunds inside Shopify, edit orders, or update inventory - actions stay with you, and action requests arrive as escalations with the customer's details attached. It also cannot answer live order-status from your Shopify admin; it answers what your shipping policy says about timelines, which covers most of the where-is-my-order volume but not a specific parcel's GPS position.

If a customer asks for something only a human with admin access can do, the right behaviour is a clean handover - which is exactly what the confidence threshold plus the person button produce.

The first week: read the gaps

After launch, the useful work moves to the escalation inbox. The questions the bot could not answer arrive as a list, and that list is your writing queue - add the missing document and the same question answers itself next time. Expect the list to be longest in week one and to shrink fast: most stores find their real top-fifteen questions differ from the fifteen they guessed, and the gap list is how you find out which. Guidance on structuring those documents is in the knowledge-base guide, and the free plan covers a small store's first month comfortably.