Your customers get answers. Your team gets their day back.
An AI assistant grounded in your own approved policies, FAQs and product documentation — so it answers like your best agent, and says “let me get someone” when it shouldn’t guess.
Answers only from approved sources · Escalates when unsure · Every conversation reviewable
The same forty questions, every week.
Where’s my order. What’s your return policy. Does this fit. How do I set it up. Your team answers them again and again, and the questions that actually need a human wait in the same queue.
Most tickets aren’t unique
A large share of inbound volume is a small set of repeated questions with settled answers.
Volume doesn’t respect your hours
Customers ask at 11pm on a Saturday and judge you by how long the answer takes.
Hiring doesn’t scale with it
Support cost rises with volume; the questions don’t get more interesting.
Give it what your team already knows.
Upload the documents you already have — return and shipping policies, product manuals, FAQ pages, warranty terms, internal answer guides. Point it at your help center or website and it reads those too. No rewriting, no retraining.
- PDFs, Word documents, spreadsheets and plain text
- Existing help center or website URLs
- Product catalog details from your connected commerce accounts
- Answers your team writes directly in ChannelHawk
You decide what it’s allowed to say. Build your own governance
ChannelHawk drafts an answer for each question it finds in your knowledge. You review, edit and approve. Approved answers are the only thing it uses in front of a customer — and anything outside them goes to a person instead of a guess.
- Approved answer library — review, edit and version every response
- Guardrails — topics it must never handle alone: refunds above a threshold, legal or medical claims, anything you name
- Tone settings — match how your brand actually talks to customers
- Escalation rules — when to hand off, to whom, and with what context attached
If the assistant doesn’t have an approved answer, it says so and brings in a human. It does not improvise.
Put it where your customers already are.
- Website widget — a chat assistant matched to your branding, live in a single embed
- Email — drafts or sends replies to your support inbox under your rules
- Help center — an answer box on your existing documentation
- Order-aware responses — connected to your commerce data, so “where’s my order” gets a real answer, not a policy link
Every channel draws on the same approved answers and the same guardrails, so the assistant behaves identically wherever a customer reaches you.
It gets better because you can see everything it does.
Every conversation is logged, searchable and reviewable. You see what was asked, what was answered, what got escalated, and — most usefully — what customers asked that you had no approved answer for. That list is your content roadmap.
- Resolution rate — how much volume the assistant handled end to end
- Escalation log — what it handed off, and why
- Knowledge gaps — recurring questions with no approved answer yet
- Customer satisfaction — feedback captured at the end of a conversation
- Full transcript history — searchable, exportable, auditable
What’s included
| Capability | Support |
|---|---|
| Approved answer library with versioning | ✓ Included |
| Document, URL and catalog ingestion | ✓ Included |
| Website chat widget with brand matching | ✓ Included |
| Order-aware answers from connected commerce data | ✓ Included |
| Escalation rules and human handoff | ✓ Included |
| Guardrails and restricted topics | ✓ Included |
| Conversation analytics and knowledge-gap reporting | ✓ Included |
| Full transcript history and export | ✓ Included |
| Included resolutions per month | 500 |
| Additional resolutions | $0.75 each |
Point it at your policies and watch what it does.
Bring your return policy and your five most common questions to a 30-minute call. We’ll load them live and you can try to break it.