Support chat
How to use it
Turn on support chat and let AI draft the first reply:
- Add the chat widget to your storefront. (Support → Storefront)
- Set your SLA target for first response. (Support → SLA)
- Let the AI agent draft replies; your team approves or edits. (Support → Inbox)
- Watch support-attributed revenue to see chat as a sales driver. (Support → Analytics)
⏱ ~15 min · 💳 Starter+ · 🎯 Faster replies, with revenue tracked
Why this matters for your business
Pre-sale questions are lost revenue when they go unanswered, and post-sale questions are churn when they go slow. A support chat on your own storefront captures both — but only if it's secure (a public storefront endpoint is an attack surface), fast (customers expect minutes, not days), and measurable (you should know which conversations drove sales). This engine is built for all three.
Security model
The chat is a public storefront surface, so its security model is deliberate:
- It's reached only through the App-Proxy-authenticated route, which derives
a trusted
shopfrom the HMAC-signed proxy request. The browser never asserts which shop it is. - The browser holds an opaque session key (an unguessable id) as its handle — never the numeric record id, never the shop.
- Every read and write is scoped by
{ shop, session }, so one visitor physically cannot read another shop's — or another visitor's — messages. - Internal agent notes are filtered out of the visitor-visible message shape; a shopper only ever sees real replies.
What you get
| Capability | What it does |
|---|---|
| Storefront widget | The shopper-facing chat on your live site |
| AI agent | Drafts replies from your catalogue/context; a human can send or edit |
| SLA tracking | Computes a first-response-due time per conversation |
| Cart follow-up | Queued follow-ups tied to a shopper's cart |
| Attribution | Credits support conversations that convert to orders |
| Analytics | Volume, response times, resolution |
How a conversation flows
AI agent + human handoff
The AI agent proposes replies grounded in your store context, so common questions (shipping, sizing, returns, product details) get an instant, accurate draft. A human can approve, edit, or take over — and because internal notes are segregated from the visitor view, your team can annotate a conversation without the customer seeing it.
SLA & attribution
- SLA — each conversation gets a first-response-due time so nothing sits unanswered past your target. Breaches are visible, not silent.
- Attribution — when a support conversation is followed by an order, the revenue is credited to support in the Revenue Hub, so you can see support as a revenue driver, not just a cost centre.
Best practices
✅ Let the AI draft, keep a human in the loop for anything nuanced — speed from AI, judgment from your team.
✅ Set SLA targets you can actually hit. A 4-hour first response you meet beats a 15-minute target you miss.
✅ Watch support attribution. Stores are consistently surprised how much revenue pre-sale chat drives — it justifies staffing it.