Shopify store support has a predictable shape. A large slice of your tickets — often 50% or more — are some variation of "where is my order?" The rest are product questions, return requests, and policy lookups that any well-trained AI agent can handle.
This guide walks through setting up an AI chatbot for a Shopify store from zero to live, including the common mistakes that cause poor deflection rates.
What your Shopify AI agent should be able to handle
Before you configure anything, be clear on the scope. A Shopify AI chatbot should handle:
- WISMO queries — "Where is my order?", "Has my order shipped?", "My tracking hasn't updated"
- Return and refund requests — explaining your policy, initiating a return, linking to the returns portal
- Product questions — sizing, compatibility, materials, availability
- Order changes — cancellations before shipment, address corrections
- Policy lookups — shipping times, international delivery, discount code rules
It should escalate to a human agent for:
- Disputes or chargebacks
- Damaged or missing items that need a replacement
- Anything where the customer is clearly frustrated after the first exchange
Step 1 — Connect your knowledge base
Start by connecting the content sources that answer your most common questions.
Your Shopify store pages: Connect your store's help centre, FAQ page, and shipping/returns policy. SupportBot crawls all linked pages — paste the root URL and it indexes everything.
Your product catalogue: For product-specific questions, you want the AI agent to have access to product descriptions, size guides, and compatibility information. The simplest approach is to sync your help centre with product-specific FAQs for your most popular items.
Your policy documents: Paste the URL of your returns policy and shipping policy pages directly. If they're embedded in your theme rather than standalone pages, export them as a document and upload the file.
Step 2 — Set up the live order lookup tool
This is the step most guides skip, and it's the one that matters most for WISMO deflection.
Without live order data, the AI agent can only answer questions about your policy — not about a specific customer's order. A customer asking "where is my order?" still needs a human if the AI agent can't look up their order status.
With SupportBot's no-code API tool builder:
- Create a new tool called "Order Lookup"
- Connect it to the Shopify Orders API (or your fulfilment provider's API)
- Define the input — the customer's order number or email address
- Define what the AI agent returns — status, carrier, tracking number, estimated delivery
Now when a customer asks about their order, the AI agent asks for their order number, queries the API mid-conversation, and returns the live status. No human needed for the vast majority of cases.
Step 3 — Configure your widget for Shopify
Install the SupportBot Shopify app from the app store, or paste the embed script into your theme's theme.liquid file just before </body>.
Widget configuration for Shopify stores:
- Position: Bottom right, above the cart icon if your theme places it there
- Launch icon: Use the chat bubble — avoid the AI agent icon, it converts lower
- Welcome message: "Hi! Ask me anything about your order or our products." — specific beats generic
- Trigger: Auto-open after 45 seconds on product pages (not homepage)
Avoid auto-opening on every page load. It's annoying. Trigger it contextually: on the cart page, on product pages for items with long delivery windows, or when a customer has been on the tracking page for more than 20 seconds.
Step 4 — Write escalation rules that actually work
Every Shopify store has edge cases the AI agent shouldn't try to handle alone. Configure these as explicit escalation triggers:
If customer message contains: "dispute" | "chargeback" | "fraud" | "stolen"
→ Escalate immediately, do not attempt to answer
If customer expresses frustration twice in the same session
→ Offer live agent or ticket creation
If order status is "returned to sender" or "delivery failed"
→ Escalate — these need manual intervention
The AI agent should never try to resolve a chargeback or a lost parcel by itself. It can acknowledge the issue and get the customer into a queue faster than the customer finding the contact form themselves.
Step 5 — Measure what matters
After two weeks live, look at these metrics:
| Metric | What it tells you |
|---|---|
| Deflection rate | % of conversations resolved without human handoff |
| CSAT score | Customer satisfaction on AI agent-handled chats |
| Unanswered questions | Gaps in your knowledge base |
| Escalation rate by topic | Where the AI agent is weakest |
A well-configured Shopify AI agent should reach 55–65% deflection within 30 days. If you're below 40%, the unanswered questions report will tell you exactly where the gaps are.
Common mistakes to avoid
Connecting too little content. The AI agent can only answer what it knows. If your FAQ page has 6 questions, expect a 6-question AI agent. Invest in documentation before launch.
Skipping the live order API. Without order lookup, you're leaving the majority of your WISMO tickets undeflected. It's the single highest-impact setup step.
Setting the AI agent to auto-open on every page. Customers don't want a chatbot popping up when they're browsing. Trigger it on high-intent pages (cart, order status, product detail) or after a meaningful dwell time.
Not monitoring the unanswered questions report. Your AI agent gets smarter every week — but only if you're plugging the gaps. Set a reminder to review it every Monday morning.
SupportBot has a one-click Shopify app install and a pre-built order lookup tool template. Start free →
