Shopify & WooCommerce support evaluation checklist

Does private order access require verified identity?

Required before launch

How to test: Try a real test order reference with the wrong email and an unverified visitor. Neither should reveal private data or confirm who owns the order.

Are stores and workspaces isolated?

Required before launch

How to test: Use two isolated test stores. Try accessing store B's order through store A's chat and agent account; both must be denied.

Can access be limited and revoked?

Required before launch

How to test: Inspect the requested store permissions. Revoke the connection and repeat a private order lookup; access must stop.

Are sensitive write actions controlled?

Required before launch

How to test: Ask for a refund, cancellation, or address change. Confirm the exact supported scope, approval requirement, authorization, and audit trail.

Can the team inspect the source of an answer?

How to test: Ask a return-policy question, inspect the source, update the policy, and check the answer again after the documented refresh time.

Does uncertainty lead to a useful handoff?

How to test: Ask about an unsupported product and an exception absent from the policy. Check that the assistant avoids invented facts and routes the question.

Are order updates and lookup failures clear?

How to test: Change a fictional order's status, then disconnect the store. Check refresh behavior and that failure produces a clear handoff rather than stale certainty.

Does a human receive context and ownership?

How to test: Ask for a person. Check the inbox owner, transcript, source context, response expectation, and the customer's ability to continue the thread.

Do the required channels work on the proposed plan?

How to test: Send a test message through every required channel. Verify delivery, replies, plan gates, provider fees, and business-hours behavior.

Can the widget be used on mobile and with a keyboard?

How to test: Test at 320px and on your checkout pages. Navigate by keyboard, inspect labels and focus, and check that the widget does not block checkout controls.

Is the full normal-month and peak-month cost clear?

How to test: Price seats, AI usage and overages, required add-ons, channel/provider fees, tax, setup, and human review for both volumes. Record the dated quote.

Can a pilot measure outcomes and support a rollback?

How to test: Define a baseline, resolution denominator, reopen rate, review time, customer feedback, data export, retention, and a tested human-support fallback.