SupportIQ — User Guide

The complete handbook for the AI support agent that resolves your Shopify tickets end-to-end.

This guide walks you through every feature in SupportIQ — what it does, when to use it, and how to get the most out of it. No technical background required. If you can run a Shopify store, you can run SupportIQ.

Table of Contents

1. What SupportIQ Does
2. Getting Started
3. The Storefront Chat Widget
4. The Inbox
5. How the AI Agent Resolves Tickets
6. Suggest Mode vs Auto Mode
7. The Guarded Actions
8. Confidence, Refund Caps & the Audit Trail
9. Grounded Answers & the Knowledge Base
10. Escalation & Brand-Voice Drafts
11. Image Attachments & Authenticity
12. Customer 360
13. Voice-of-Customer & Benchmarks
14. Macros
15. Settings
16. Plans & Billing
17. Troubleshooting & FAQ
18. Getting Help
19. Glossary

What SupportIQ Does

Most "AI" support apps look up an order, draft a reply, and hand it back to you to send. They stop at the line that actually matters: issuing the refund, cancelling the order, fixing the address. SupportIQ crosses that line. It resolves the whole ticket.

A customer asks "where's my order?", "cancel this," or "I need a refund" in the storefront chat. SupportIQ understands the intent, pulls the real order and fulfillment details from your store, answers from your policies, and — when you let it — takes the action directly inside Shopify. Then it shows you exactly what it did.

The "AI" in SupportIQ maps to real capabilities:

• It acts, not just suggests — issues refunds, cancels and edits orders, updates shipping addresses, sends live tracking, starts reorders, and applies discounts directly in Shopify.
• It answers from your real data — your policies, FAQs, product details, and order data — and cites where the answer came from, instead of inventing a return window you don't offer.
• It knows when to stop — if it isn't confident, the refund is over your limit, the customer asks for a person, or a message hits a sensitive (legal/fraud) topic, it escalates to your inbox with a briefing and a ready-to-send draft in your brand voice.
• It learns what your customers are saying — clusters every conversation into counted themes so your inbox becomes a product and ops radar.

You stay in control of how far it goes. Every store starts in suggest-only mode — the agent proposes, you approve — and you graduate it to autonomous actions when you're ready.

Who is SupportIQ for?

• High-volume Shopify stores drowning in "where's my order?" (WISMO), refund, cancellation, and address-change tickets.
• Lean teams that want the repetitive volume handled automatically so people focus on the nuanced cases.
• Merchants burned by per-resolution pricing who want a flat, predictable bill that doesn't balloon on a viral day.
• Stores that take photos in support — damaged-item claims, authenticity disputes — who want the agent to see the image and act safely.

The flat-pricing wedge

The agents that actually act almost all bill you per resolution — usually around a dollar each — so the better they work, the bigger your bill, and one busy month can double it. SupportIQ is flat. You pick a plan with a hard monthly cap and pay the same price whether it's your quietest month or your biggest sale ever. And human-escalated conversations never count against your cap — you only spend it when the AI does the work.

Getting Started

1. Install the App

Join the SupportIQ waitlist or request early access from Kernvel while the app is in private launch. The app is free to start — you only pay if you upgrade to a paid plan. Once your store is approved, you land in the SupportIQ dashboard inside your Shopify admin.

2. The Onboarding Wizard

The first time you open SupportIQ, a short setup wizard walks you through the essentials:

• Connect your store — SupportIQ reads your orders and fulfillment data so it can answer "where's my order?" and take order actions.
• Import your knowledge — it pulls in your existing Shopify policies and pages (refund policy, shipping policy, FAQ, etc.) so the agent is grounded from day one.
• Enable the chat widget — turn on the storefront chat so customers can reach the agent.
• Set the agent persona — give it a display name, a greeting, and a tone.

You can't get lost: SupportIQ keeps you in the wizard until the basics are done, then drops you on the dashboard.

3. Enable the Storefront Chat Widget

In your Shopify admin, go to Online Store → Themes → Customize → App Embeds, find the SupportIQ chat widget in the list, and toggle it ON. This is what makes the chat launcher appear on your storefront. One toggle, no code, every theme.

4. Start in Suggest Mode

Out of the box, SupportIQ is in suggest-only mode: it answers questions and proposes actions, but never issues a refund or cancels an order on its own. Watch how it responds for a while. When you're comfortable, switch it to auto mode (on a paid plan) and let it close tickets autonomously, within the caps you set.

5. That's It

The agent is now answering customers in the storefront chat, and anything it can't handle lands in your inbox already understood. Setup is minutes, not days.

The Storefront Chat Widget

The storefront chat widget is how your customers reach SupportIQ. It's a Shopify App Embed block (the support-widget theme extension), so it works on every Shopify theme — Dawn, Craft, Refresh, and third-party themes — with no code.

What customers see

• A chat launcher in the corner of your storefront.
• A conversation window where they type their question in plain language.
• The agent's replies, with order details, tracking, and answers pulled live from your store.
• The ability to attach a photo (for example, a picture of a damaged item) when relevant.

What it can do, right there in the chat

• Answer questions from your policies and pages ("what's your return window?", "do you ship to Canada?").
• Look up an order by its number and report status and tracking ("where's my order #1042?").
• On a paid plan in auto mode, take the action — issue the refund, cancel the order, change the address, start a reorder, apply a discount — within your caps.
• Hand off to a human when it should, so the customer is never stuck with a wrong answer.

Customization

You control how the widget looks and sounds:

• Accent color and launcher text — set on the Settings screen, so the chat button matches your brand.
• Display name, greeting, and tone (friendly, formal, or playful) — set on the Agent screen.

The widget picks all of these up so the chat feels like part of your store, not a generic bot. Customers can also attach a photo and, after a conversation is resolved, rate it.

Note on theme password mode: if your store is still password-protected (a brand-new store in "coming soon" mode), the widget needs your storefront to be live to fully connect. Once you remove the password gate, the widget detects itself automatically — the onboarding wizard shows "Widget detected ✓".

The Inbox

The Inbox (Conversations) is your home base — every support conversation across your connected channels, in one place. SupportIQ handles two channels today:

• Storefront chat — the chat widget on your store (always on once enabled).
• Email — an optional support email channel powered by SendGrid Inbound Parse. SupportIQ gives you a unique inbound address; you forward your support@ mailbox to it, and incoming emails become conversations the same agent works on, with replies emailed back. (This channel needs a one-time go-live step — forwarding plus a mail record — described in Settings.)

The conversation list

Each row shows a conversation at a glance:

• The customer and the latest message.
• Status — where the conversation is in its lifecycle (see below).
• A one-line action summary — a plain-English summary of the most significant thing the agent did on that conversation (for example, "refund issued," "order cancellation — awaiting approval," "handoff"). It's privacy-first: the row never shows raw customer data, just the action and its state.

Conversation status

Every conversation moves through a clear lifecycle:

• Open — new, not yet handled.
• AI handling — the agent is working on it.
• Needs human — escalated; it's waiting for you.
• Human handling — a person on your team has picked it up.
• Resolved — closed out. If the customer replies, it reopens automatically.
• Closed — terminal.

The conversation detail view

Open any conversation to see the full thread plus everything the agent did:

• The full message history — customer and agent, in order.
• The agent's actions — the audit trail of what it did: which order, which action, and why. (More in the Audit Trail.)
• Any attachments — photos the customer shared, with the authenticity verdict where applicable.
• The Customer 360 panel — a live, privacy-safe profile of the customer (more in Customer 360).

What you can do from a conversation

• Reply — type a message to the customer yourself.
• Use Copilot — a writing assistant built into the reply bar. It can draft a reply for you, summarize a long thread so you catch up fast, rewrite what you've written in a different tone, or translate it (multilingual is a Pro-and-above feature). You stay the author; Copilot just speeds you up.
• Insert a macro — drop in a saved reply for the things you answer the same way every time (see Macros).
• Assign — hand the conversation to a teammate.
• Resolve — mark it done.
• Reopen — bring a resolved conversation back.
• Approve a held action — when the agent proposes a money-moving action that's over your auto limit (or you're still in suggest mode), it parks the action as awaiting approval. You review it and approve it with one click, and SupportIQ executes it for real — or you decline and handle it your way.

Customer satisfaction (CSAT)

After a conversation is resolved, customers can rate it (1–5 stars) and leave a comment. Those ratings roll up into your dashboard and insights so you can see how the agent — and your team — are actually landing with customers, and read the recent comments verbatim.

How the AI Agent Resolves Tickets

When a message comes in, the agent runs a short, guarded loop:

1. Understand the intent. What does the customer actually want — a status check, a cancellation, a refund, a policy answer?
2. Gather the facts. It looks up the real order, fulfillment, and customer details from your store, and searches your knowledge base for the relevant policy.
3. Decide what to do. It either answers, takes an action, or escalates — based on its confidence and your rules.
4. Act (or propose). In auto mode, within your caps, it performs the action directly in Shopify. Otherwise it proposes the action for your approval, or escalates.
5. Record it. Every action is written to the conversation's audit trail — what it did, on which order, and why.

How a turn ends

Every agent turn is honestly labelled with how it ended:

• Auto-resolved — the agent took an action that closed the issue.
• Answered — it answered the question (no action needed).
• Escalated — it handed the conversation to a human.
• Pending — it's mid-flight or waiting on something (like your approval, or a photo).

These labels feed your dashboard's deflection and auto-resolution numbers, so you can see exactly how much real work the agent is doing.

Preview it safely first

Before you trust the agent on live customers, you can test it in a sandbox. From the Agent screen, type a sample customer message and watch the agent respond — its reply, its proposed actions, its citations, and its confidence — without it touching anything. Sandbox turns are clearly flagged, never counted against your plan, and never written to your inbox. It's the no-risk way to tune tone and behavior.

Suggest Mode vs Auto Mode

SupportIQ has two modes, and you choose which one each store runs in.

Suggest mode (the default)

The agent answers and proposes, but never acts on its own. It looks up the order, drafts the reply, and recommends the action — then leaves the action for you to approve. This is where every store starts. Use it to build confidence: you see exactly what the agent would do before you let it do anything.

Auto mode

The agent takes the action directly — issues the refund, cancels the order, changes the address — without waiting for you, as long as the action is within your caps and clears the confidence bar. The hard cases still escalate; the clean ones close themselves.

Autonomous actions are a paid feature. On the Free plan the agent is suggest-only (answer + propose) — turning on autonomous actions is the headline reason to upgrade to Starter or above. Even in auto mode, you can keep any individual action on a short leash with the caps and allow-list below.

You flip the mode in the Agent screen. You can switch back to suggest at any time.

The Guarded Actions

The agent works through a fixed, well-defined set of tools — there's no open-ended "the AI can do anything." Some are read-only (safe to run any time); some mutate your store (and are gated behind confidence, caps, the autonomous-actions tier flag, and auto mode).

Read-only actions (always available)

Mutating actions (gated)

You decide which of these mutating actions the agent is even allowed to attempt. In the Agent screen you can allow-list exactly the actions it may take (for example, allow refunds and tracking but never auto-cancel). Anything off the list is simply never attempted — the agent escalates instead.

Orders are matched by number, the way customers refer to them. A customer only ever knows their order number ("#1042"). SupportIQ resolves that to the right order before any action, so lookups, refunds, cancellations, and tracking always hit the correct order.

A shopper can only act on their own order. Account-changing actions (a refund, a cancellation, an address edit) require the shopper to be signed in to your storefront and to actually own the order — a visitor can't refund or cancel someone else's order by guessing its number. When that ownership can't be confirmed, the action is parked for your approval instead of executed. Order lookups are identity-checked too, and personal details are masked in what the agent surfaces.

Confidence, Refund Caps & the Audit Trail

This is the safety core that makes autonomy trustworthy.

Confidence gating — never act unsure

The agent only executes a store-changing action when its confidence clears a floor. Below that bar it does not guess and act — it escalates to a human instead. "Never act unsure" is the default behavior, not an option you have to find.

Refund caps — a hard ceiling on what it can give away

You set the maximum refund the agent may issue on its own. Each plan has a ceiling, and you can set a lower per-store override (you can tighten the cap below your plan's limit, but never raise it above it):

A refund above your cap doesn't just fail silently — the agent parks it as "awaiting approval" in your inbox, so you can approve the larger refund yourself with one click, or decline it. The agent also has matching caps on discounts (a max percentage and max dollar amount) and on reorder drafts (a max number of items and a max value), so it can never quietly give away more than you've allowed.

Your own guardrail rules

Beyond the caps, you can write plain-language rules the agent must always obey — for example, "never cancel an order that's already shipped," or "always ask for a photo before refunding a damaged item." These are a safety feature available on every plan (no upgrade required), and the agent honours them on every turn.

The visible audit trail

Every action the agent takes — proposed, awaiting approval, executed, or failed — is recorded on the conversation as an AgentAction: what it did, on which order, and why. Nothing is a black box. You can open any conversation and see precisely how the agent handled it, which is exactly what you want when you're deciding whether to trust it with more autonomy.

The runaway-spend backstop

Each store also has a hard monthly AI-spend ceiling. If usage ever crosses it — a traffic spike, an attempted prompt-injection, anything — the agent automatically falls back to a safe mode (answer and escalate, no autonomous actions) rather than running up your bill. Combined with flat pricing, this means there is no scenario where a busy month produces a surprise charge.

Grounded Answers & the Knowledge Base

SupportIQ answers from your content, not a generic guess off the internet. If you don't offer a 90-day return window, it will never invent one — and if it doesn't have the answer, it tells the customer it's getting a human rather than making something up.

How grounding works

The agent searches your knowledge base for the passages most relevant to the customer's question, then answers from those passages and cites the source it used. Behind the scenes this is semantic search — it matches on meaning, not just keywords — so a customer asking "can I send it back?" still finds your "Returns & Refunds" policy.

The Knowledge screen

The Knowledge screen is where you manage what the agent knows:

• Imported Shopify content — on setup, SupportIQ pulls in your store's policies and pages so the agent is grounded immediately.
• Articles you add — write your own FAQ-style articles for anything not already on your store (sizing guidance, care instructions, warranty terms).
• Uploaded documents — upload a document (such as a PDF policy or guide) and SupportIQ extracts the text and indexes it.

Each plan includes a generous article allowance (from 25 articles on Free up to 10,000 on Enterprise), and there's a per-document size limit so one enormous file can't crowd everything else out.

It stays current

When you change an article, SupportIQ re-indexes your knowledge base so the agent answers from the latest version. If a customer asks something your knowledge base can't answer confidently, the agent escalates — and that gap shows up in your insights as a knowledge-base gap you can fill (more in Voice-of-Customer).

Escalation & Brand-Voice Drafts

The whole point of escalation is that the hard tickets — the angry ones, the over-policy ones, the legally sensitive ones — land in your inbox already understood, not as a cold message you have to decode from scratch.

When the agent escalates

The agent hands a conversation to a human when:

• It isn't confident enough to act or answer.
• A refund (or other action) is over your limit.
• The customer asks for a person — a "can I talk to a human?" routes straight to your inbox.
• A message contains legal or fraud-sensitive language — terms like "lawyer," "chargeback," or "fraud" are detected automatically and always route to a human.
• The customer hits one of your escalation triggers — you can add your own keywords (anything specific to your business) that always route to a human.
• The agent repeatedly fails to complete an action.

The agent also reads each conversation's sentiment (positive, neutral, or negative) and attaches it to the handoff, so when a conversation does escalate you instantly see the customer's mood — and negative sentiment is exactly what your Insights surface as a trend.

What you get when it escalates

• A one-line "why this needs you" — a short briefing so you're instantly up to speed.
• A ready-to-send draft reply — written in your brand voice, so you approve or lightly edit instead of authoring from a blank page.

Teaching it your voice

You can paste a brand-voice sample — a few of your own past replies — and SupportIQ uses it so its escalation drafts sound like you, not a generic bot. Brand-voice drafts are included on Starter and above.

Where escalations go

Escalations appear in your Inbox as "needs human," and you can also set an escalation email so your team gets notified the moment a conversation needs a person.

Image Attachments & Authenticity

Some support tickets only make sense with a picture — a damaged item, a wrong product, an authenticity dispute. SupportIQ lets customers attach photos in the chat, and the agent can actually see them.

Photo attachments

Customers can attach a photo to a conversation. The image is stored securely, shown in the inbox, and the agent can reference it when reasoning about the ticket. Each plan includes a monthly image allowance.

Authenticity check (Starter and above)

When a photo matters to a money decision, SupportIQ runs an authenticity check before the agent acts on it. It looks for signs that an image is AI-generated or reused rather than a genuine, original photo, using a combination of signals:

• AI-generation provenance — scanning for the content-credential markers some AI tools embed.
• A vision second opinion — the agent's own visual judgment of whether the image looks generated.
• Photo metadata (EXIF) — the camera/source fingerprint a real photo usually carries.
• Reuse detection — checking whether the same image has been seen before.
• Optional reverse-image search — if you've connected it, an extra check against the web.

Crucially, when an image is part of a refund or other money-moving request, the agent won't take the money action until the authenticity check has run — and if the photo looks suspect, it escalates to you rather than acting. The agent's job here is to see and judge, not to silently approve.

Image intelligence & defect clustering (Pro and above)

On higher plans, SupportIQ goes a step further and turns support photos into product intelligence. It groups visually similar images together so recurring problems surface as clusters — for example, "fourteen customers sent photos of the same seam tearing on the Midnight Hoodie." That's a defect signal you'd otherwise never see, pulled straight from your support inbox.

Customer 360

When you open a conversation, SupportIQ assembles a live, privacy-safe profile of the customer right in the inbox — so you (and the agent) resolve the ticket with the full picture instead of asking the customer to repeat themselves.

What's in the profile

Customer 360 composes a view from your store and, when you have other Kernvel apps installed, from those too:

• Support history — past conversations with this customer.
• Orders — their order history and status.
• Wishlist — what they've saved (from ListIQ).
• Membership — their tier and perks (from GateIQ).
• Return risk — a coarse, advisory signal (from ReturnsIQ).
• Auction activity — their bidding history (from AuctionAI).

Privacy by design

Customer 360 is assembled on demand and stored as no durable dossier — it's a live view composed when you open the conversation, not a profile sitting in a database. The sister apps share only the minimum needed (counts, flags, tiers — never raw customer lists), and you control every connection.

For the full picture of how the cross-app context works — Customer 360, Product 360, and how membership and return context enrich support — see the Integrations & Connections guide.

Voice-of-Customer & Benchmarks

Your support inbox is the most honest product feedback you'll ever get. SupportIQ turns it into a radar — automatically, from real counted conversations, never a hand-wavy "mostly positive."

The always-on dashboard (Free)

Every plan, including Free, gets a value dashboard on the home screen:

• Deflection funnel — how many conversations the AI handled versus escalated.
• Auto-resolution rate — the share the AI resolved on its own, with a 4-week trend.
• Agent actions performed — a count of the real actions taken (refunds, cancels, etc.).
• Estimated time saved — the human hours the agent gave you back.
• Top customer intents — what people are actually contacting you about.
• Escalation reasons — why the agent handed conversations to a human.

Voice-of-Customer insights (Pro and above)

On Pro and above, SupportIQ clusters your conversations into named, counted themes — "23 customers this week asked about the Midnight Hoodie being out of stock; 11 said shipping to Texas was slow." Each theme carries a sentiment, a week-over-week trend, sample quotes, and a suggested action (restock, content, shipping, product, or policy), so you know not just what customers are saying but what to do about it. The same engine flags knowledge-base gaps (questions your articles can't answer) so you know exactly what to write next, plus a sentiment trend so you can see customer mood moving over time.

AI quality monitoring (also Pro and above) automatically scores a sample of the agent's resolutions and flags any that look off — a likely wrong answer, a missed escalation, an off-policy action — and lists them for you to review. It's an honest self-check on the agent's work, so you can trust the autonomy and catch the rare miss.

Cross-shop benchmarks (Scale and above)

On Scale and above, SupportIQ shows you how your support stacks up against comparable stores — anonymously and in aggregate. Benchmarks are only ever shown when enough stores are in a bucket to keep every store anonymous (a strict minimum-group rule), so you get a useful comparison without anyone's individual data being exposed.

Macros

Macros are saved, reusable replies for the situations you handle the same way every time — "here's how to start a return," "our holiday shipping cutoffs," and so on. Instead of retyping the same answer, you (or your team) drop in a macro and send. They keep your human replies fast and consistent, and they pair naturally with escalation: when the agent hands you a conversation, a macro often finishes it in one click.

Macros support placeholders that fill in automatically for the conversation you're in, so a saved reply still reads personally:

• {{customer_name}} — the customer's name
• {{order_number}} — their order number
• {{order_status}} — the order's current status
• {{shop_name}} — your store name
• {{agent_name}} — the agent's display name

You create, edit, and toggle macros active or inactive from the Macros screen.

Settings

The Settings area is where you configure how SupportIQ behaves. The most important controls live on the Agent screen; broader store settings live on Settings.

Agent persona

• Display name — what the agent calls itself in the chat (e.g., "Ava").
• Greeting — the opening message customers see.
• Tone — friendly, formal, or playful.

Autonomy & safety

• Mode — suggest (propose only) or auto (act directly).
• Allowed actions — the allow-list of which store-changing actions the agent may take.
• Auto-refund cap — the most the agent may refund on its own (within your plan's ceiling).
• Guardrail rules — plain-language policies the agent must always obey.

Escalation

• Escalation keywords — words or phrases that always route a conversation to a human.
• Escalation email — where your team is notified when a conversation needs a person.
• Brand-voice sample — your own past replies, so escalation drafts sound like you.
• Proactive prevention (Pro and above) — when an order's fulfillment is delayed or fails, SupportIQ can get ahead of the inevitable "where's my order?" by preparing a heads-up draft, so you reach out before the customer has to ask.

Storefront widget

• Enable the widget and confirm it's live on your storefront (the screen shows a "detected" status).
• Accent color and launcher text — style the chat button to match your store.
• A deep link to the theme editor so you can toggle the App Embed in one place.

Email support channel

• Your unique inbound support address (with a copy button) — forward your support@ mailbox to it to turn email into agent-handled conversations.
• A status indicator that tells you when the channel is fully live. (Going live needs a one-time forwarding + mail-record setup; SupportIQ shows you exactly what's pending.)

Knowledge

Manage articles, import Shopify policies and pages, and upload documents (see the Knowledge Base).

Connections

Control Customer 360 context from your other Kernvel apps and see which Product 360 sources are connected — see the Integrations guide.

Billing

View your plan, usage, and limits; upgrade, downgrade, or cancel (see below).

Plans & Billing

SupportIQ is flat-priced with hard caps — never "unlimited," and never a charge per resolution. The primary metric is AI resolutions per month (conversations the AI handled). Conversations escalated to a human do not count against your cap. Starter, Pro, and Scale each include a 14-day free trial (Enterprise is sales-assisted, with trial and onboarding handled as part of your launch plan), and you can save up to ~17% with annual billing (pay 10× the monthly price, get 2 months free).

A resolution is a conversation the AI actually handled for you — answered or auto-resolved. Human-escalated conversations don't count, by design.

Free — $0/month

• 50 AI resolutions / month
• Storefront chat widget
• Grounded answers from your data, with citations
• Suggested replies (answer-only — the agent proposes, you act)
• 25 knowledge articles
• The always-on value dashboard

Good for trying SupportIQ on real customers with zero risk.

Starter — $19/month

• 250 AI resolutions / month
• Autonomous actions turn on (refunds, cancels, address changes, reorders, discounts)
• Auto-refund cap up to $50
• Brand-voice drafts & smart escalation
• Image attachments + authenticity check
• 100 knowledge articles

Good for a small store ready to let the agent close clean tickets on its own.

Pro — $49/month (recommended)

• 1,000 AI resolutions / month
• Everything in Starter
• Auto-refund cap up to $150
• Voice-of-Customer insights + sentiment & knowledge-gap analysis
• Image intelligence (defect clustering)
• Proactive WISMO prevention + multilingual answers
• AI quality auto-scoring
• 500 knowledge articles

Good for active stores that want the agent acting and the inbox working as a product radar.

Scale — $149/month

• 5,000 AI resolutions / month
• Everything in Pro
• Auto-refund cap up to $500
• Cross-shop benchmarks (anonymous, aggregate)
• Highest autonomy caps + priority processing
• 2,000 knowledge articles

For high-volume merchants running support as a core function.

Enterprise — $399/month

• 20,000 AI resolutions / month
• Everything in Scale
• Configurable refund & action caps (no fixed ceiling)
• Bring-your-own-key option + custom guardrails
• 10,000 knowledge articles
• SLA & dedicated onboarding

Switching plans

Change plans any time from Billing. Upgrades and downgrades go through Shopify's normal subscription billing, and your caps adjust to the new plan. Cancel any time — you'll move to the Free plan at the end of your current billing period, and no data is deleted.

Troubleshooting & FAQ

"The chat widget isn't showing on my storefront."

Check two things in order:

1. Is the SupportIQ App Embed enabled in your theme? (Online Store → Themes → Customize → App Embeds → toggle SupportIQ ON.)
2. Is your storefront live (not password-protected)? A brand-new store in "coming soon" mode can stop the widget from fully connecting. Once you remove the password gate, it detects itself — onboarding shows "Widget detected ✓".

"The agent answered but didn't take the action."

Most likely one of these by design:

• You're in suggest mode — the agent proposes; you approve. Switch to auto (on a paid plan) to let it act.
• The action isn't on your allow-list — add it in the Agent screen.
• A refund was over your cap — it's parked as "awaiting approval" in your inbox; approve it there.
• The agent wasn't confident enough — it escalated instead of guessing.

"Will my bill explode in a busy month?"

No. SupportIQ is flat-priced with hard caps — one predictable price whether it's your quietest month or your biggest sale. There's never a charge per resolution, and human-escalated conversations don't count against your cap. A hard monthly AI-spend ceiling also means a runaway agent can never quietly run up your bill.

"What if the AI isn't sure, or gets something wrong?"

It escalates instead of guessing. The agent answers only from your real policies and order data, and when something falls outside what it confidently knows — or a refund is over your limit, the customer asks for a person, or a message hits a sensitive (legal/fraud) topic — it stops, routes the conversation to your inbox, and hands it to you already understood, with a one-line summary and a brand-voice draft.

"How do I stay in control of what it can do?"

Every store starts in suggest-only mode. You set the maximum refund the agent may issue on its own, allow-list exactly which actions it may take, and write plain-language rules like "never cancel an order that's already shipped." Everything it does is logged in the audit trail, so autonomy is always visible.

"Where do its answers come from?"

From your store — your policies, FAQs, product details, and live order data — and it cites the source. On setup it imports your Shopify policies and pages. If you don't offer a 90-day return window, it won't invent one; if it doesn't have the answer, it gets a human.

"Does it replace my support team?"

No — it removes the repetitive volume so your team handles only what genuinely needs a human. SupportIQ is built as AI-assist plus human escalation, never AI-only.

"Which Shopify themes does it work with?"

All of them. The chat widget uses Shopify App Embed blocks, so it works with every theme — Dawn, Craft, Refresh, Debut, and third-party themes — with no code changes.

"Is my customers' data safe and private?"

Yes. Access tokens are encrypted at rest, and each store's data is fully isolated. SupportIQ honours Shopify's data-request, customer-redact, and shop-redact requests automatically, with configurable retention and a Data Processing Agreement on request. Customer 360 is assembled on demand and kept as no durable dossier.

"Can I refund a customer myself instead of letting the agent do it?"

Yes. You can reply, take over, and approve or decline any proposed action right from the conversation. And you can always refund through Shopify directly the way you normally would.

Getting Help

In-App Support

Reach out from inside SupportIQ any time — your plan determines response speed (see below).

Email Support

For longer or sensitive questions, email support@kernvellabs.com. Include:

• Your store URL
• The conversation or feature you're asking about
• Screenshots if relevant

Help by Plan

• Free — community/email support (48-hour target).
• Starter — email support (24-hour target).
• Pro / Scale — priority support (24-hour / 12-hour targets).
• Enterprise — SLA with a dedicated onboarding contact (4-hour target).

Documentation

You're reading it. This guide is updated as new features ship — check back when you upgrade plans or notice something new in the app. For cross-app capabilities, see the Integrations & Connections guide.

Glossary

Agent — the AI that reads each customer message, gathers facts, and either answers, acts, or escalates.

AgentAction / Audit Trail — the record on each conversation of every action the agent took: what it did, on which order, and why.

Auto mode — the agent takes store-changing actions directly (within your caps), without waiting for approval.

Auto-refund cap — the most the agent may refund on its own. Each plan has a ceiling; you can set a lower per-store override.

Awaiting approval — a money-moving action the agent has proposed but parked for your one-click approval (because it's over your cap or you're in suggest mode).

Confidence gate — the agent only acts when it's confident enough; below the bar it escalates instead of guessing.

Deflection — a conversation the AI handled so it never needed a human.

Escalation — handing a conversation to a person, with a briefing and a brand-voice draft.

Grounded answer — an answer drawn from your own policies, pages, and order data, with the source cited.

Knowledge base — your policies, pages, articles, and uploaded documents that the agent answers from.

Macro — a saved, reusable reply for situations you handle the same way every time.

Mutating action — an action that changes your store (refund, cancel, edit address, reorder, discount); gated behind confidence, caps, and auto mode.

Resolution — a conversation the AI handled (answered or auto-resolved). The metric your plan cap measures. Human-escalated conversations don't count.

Suggest mode — the agent answers and proposes actions, but never acts on its own. The default for every new store.

Voice-of-Customer (VoC) — counted themes the agent clusters from your conversations, turning your inbox into a product and ops radar.

WISMO — "Where Is My Order?" — the single most common support question, and the agent's bread and butter.

Document version 1.0 · Last updated June 2026 · Questions? support@kernvellabs.com