WhatsApp is where they message. Hay.chat acts in the systems behind it.
The integrations you've connected don't care which channel the question came from. If Hay.chat can refund from your web chat, it can refund from a WhatsApp message.
WhatsApp AI customer service only earns its keep if it can act. Hay.chat reads the message, pulls what it needs from your stack, and gets the job done. Often before the customer types again.
Messages are end-to-end encrypted. This business uses a secure service to manage the chat.
When a customer messages you on WhatsApp, they expect the same experience they get messaging anyone else. Fast, conversational, done. Most AI tools don't deliver that. They answer a question, push a knowledge base article, or collect details and promise a follow-up.
Underneath, it's still just a form with a chat bubble on top.
Hay.chat treats WhatsApp as a place where things get done. Orders get tracked. Refunds happen inside the chat. The customer never has to leave WhatsApp to get a result.
2B+
WhatsApp users worldwide
177
Actions Hay.chat can take across your stack
9
Deep integrations powering them
WhatsApp messages are short, informal, and often arrive three in a row. Hay.chat reads the whole thread, works out what the customer actually wants, and picks up details like an order number or an email even when they're buried mid-sentence.
Integration Hay.chat NLU
Your customers are already on WhatsApp. Meet them there properly.
Start Your Free TrialThe integrations you've connected don't care which channel the question came from. If Hay.chat can refund from your web chat, it can refund from a WhatsApp message.
Drag and drop your return policy, shipping times, and product guides. Hay.chat accepts PDF, DOCX, and HTML, or crawls your help centre directly.
Categorise documents, set visibility, and toggle which ones your AI agent can access. Internal margin sheets stay internal.
Update documents anytime. Hay.chat always uses the latest version, so a policy change reaches WhatsApp the moment you save it.
The dashboard below is filled with sample data so you can see the shape of the reporting. These are not customer results.
Hay hands a conversation to a person below its confidence threshold, which defaults to 75% and is yours to move. Every decision it made on the way is in the log.
One playbook, written in plain English, deciding what Hay.chat does with a WhatsApp thread from first message to handoff.
1 This is WhatsApp, so keep it to a line or two. Ask one question at a time. People are typing with a thumb.
2 If they're asking about an order, get the email or order number, then look it up with Shopify · Get Order and reply with the carrier and the tracking link.
3 For a return, check it against Returns policy and refund with Stripe · Refund Payment up to £100. Anything larger goes to a person.
If they send a photo or a voice note, say plainly that you can only read text right now, and ask them to describe it. Don't pretend you looked at it.
4 When you hand over, stay in this thread. Open Zendesk · Create Ticket with the whole conversation attached, and tell the customer a person is picking it up.
Channel-specific by design. The same trigger can run a different playbook on WhatsApp than on email. Shorter replies, different tone, same underlying logic.
Written for the medium. WhatsApp is a phone keyboard. The playbook says so, and Hay.chat keeps its replies to a couple of lines instead of pasting a policy page.
Type @ to reach your stack. The body is a document, not a flowchart. Type @ and pick Get Order, and Hay pulls the order straight from Shopify.
Guardrails, not guesswork. A refund cap you set, checked against the policy document, before any money moves. Above the cap it goes to a person.
Handoff stays in WhatsApp. Escalation is a line you write. The customer keeps the same thread, and your agent arrives with everything already looked up.
AI handles a lot of WhatsApp support. It doesn't have to handle all of it. Hay.chat routes the conversations that need a human to a human, with the full chat and customer context already loaded. You decide where the line sits.
Escalation rules you control. Set the conditions for handoff yourself. A refund above your threshold. Sentiment going sour. A customer asking for a human. The rules sit in your playbooks, where you can read and edit them in plain language.
Handoff keeps the chat. When Hay.chat escalates, the conversation continues in WhatsApp with a human picking up where the AI left off. The customer doesn't have to start again on email or in a portal.
Action guardrails. Refund caps, eligibility checks, scopes on write access. Hay.chat only takes the actions you've authorised, and only when the policy says it can.
Audit trail. Every action Hay.chat takes gets logged. The reasoning behind each decision is reviewable. If something goes sideways, you can see exactly what happened and why.
EU data hosting. Conversation data sits in the EU on DigitalOcean. GDPR-aligned by default.
The aim is to clear the repetitive tickets off your team's plate. Honestly, that's the whole point. They get more time for the conversations that need a person.
Connect WhatsApp through Twilio, point Hay.chat at your store and your helpdesk, and let it start closing threads instead of opening tickets.