WhatsApp is already the preferred support channel for 70% of users in Latin America and growing fast in the US too — ahead of phone and email in many demographics. But offering WhatsApp Business support at scale — with multiple agents, metrics, automation and AI — requires more than the free app. This guide shows how to set up professional WhatsApp Business support, from zero to mature operation.
Why WhatsApp works for support
- Read rate: 98% (vs 25-30% for email)
- Customer already has it installed (~80% of US adults, 95%+ in LATAM)
- Reply in minutes becomes norm, not exception
- Permanent history — customer returns and continues the conversation
- Rich media — images, videos, PDFs without friction
The counterpart: expectation of fast reply. A customer waiting more than an hour on WhatsApp rates you poorly.
The 3 levels of WhatsApp support
Level 1: WhatsApp Business app (up to ~50 conversations/day)
Good for:
- Businesses starting out
- Micro-companies, brick-and-mortar
- 1 agent, 1 phone
Features:
- Greeting and away auto-messages
- Labels (Pending, Resolved, etc.)
- Quick replies
- Catalog
Main limitation: one phone only. Doesn't scale for multiple agents.
Level 2: API + multi-agent dashboard
Good for:
- 100-1,000 conversations/day
- 2-10 agents
- CRM integration
How it works: the WhatsApp Business API connects your number to a dashboard where multiple agents work in parallel. Each sees which conversations are theirs, which are free, which are "waiting".
Common platforms: Zavu (with native dashboard), Twilio Flex, MessageBird Inbox, ZenDesk with WhatsApp integration.
Level 3: API + automation + AI (any volume)
Good for:
- 1,000+ conversations/day
- Need 24/7 support
- Multiple languages
Combines:
- AI agent for triage and FAQs (answers the repetitive questions)
- Smart routing by question type
- Shared inbox where your team reads and answers any conversation
- Webhooks for CRM, e-commerce, ERP
How to set up WhatsApp support in 5 steps
1. Decide the level
Practical math: daily volume × average response time.
- Up to 50 conv/day → app
- 50-300 conv/day → API with dashboard
- 300+ → API + AI
2. Get API access
For levels 2 or 3, create free Zavu account, connect your Meta Business Manager and verify the number. All in ~30 minutes.
3. Create greeting and FAQ response templates
Customer expects first reply in < 1 minute. Solutions:
Instant greeting template (responds in seconds)
TypeScriptimport Zavudev from "@zavudev/sdk" const zavu = new Zavudev({ apiKey: process.env.ZAVUDEV_API_KEY }) // Webhook when receiving message if (event.type === "message.inbound") { await zavu.messages.send({ to: event.data.from, channel: "whatsapp", text: "Hi! We got your message. An agent will reply within 15 minutes." }) }
Keyword-based replies (resolves 30-40% without humans)
TypeScriptconst text = event.data.text.toLowerCase() const replies = { "tracking|where": "To track your order, go to zavu.dev/tracking with your order ID.", "hours|time": "We're open Monday to Friday, 9am to 6pm.", "return|refund": "Returns within 7 days. See: zavu.dev/returns", } for (const [keywords, resp] of Object.entries(replies)) { if (text.match(new RegExp(keywords))) { await zavu.messages.send({ to: event.data.from, channel: "whatsapp", text: resp }) return } } // No match → leave it for your team in the inbox
4. Configure routing by type
Customer picks the topic via interactive buttons:
TypeScriptawait zavu.messages.send({ to: event.data.from, channel: "whatsapp", messageType: "buttons", text: "How can we help?", content: { buttons: [ { id: "sales", title: "Buy" }, { id: "support", title: "Support" }, { id: "billing", title: "Billing" } ] } }) // When customer clicks, route to correct queue const reply = event.data.content?.interactiveReply // a tap arrives as message.inbound if (event.type === "message.inbound" && reply) { const queue = reply.id // "sales" | "support" | "billing" await routeToQueue(event.data.from, queue) }
5. Metrics that matter
Measure:
- Average first response time — target: < 5 min
- Average resolution time — target: < 30 min
- AI-resolved conversations — more = cheaper
- Post-support NPS — target: > 70
- Abandonment rate (customer gave up) — target: < 5%
Compute them from your conversation and message data (GET /v1/conversations, GET /v1/messages); the dashboard shows message delivery and what the AI agent answered.
AI support (that actually works)
Zavu's AI agents use RAG (Retrieval Augmented Generation) — they answer from your knowledge base:
TypeScriptconst senderId = "sender_abc123" // 1. Create the knowledge base and add your documents (text content) const { knowledgeBase } = await zavu.senders.agent.knowledgeBases.create(senderId, { name: "Support docs" }) await zavu.senders.agent.knowledgeBases.documents.create(knowledgeBase.id, { senderId, title: "Return policy", content: "Returns are accepted within 7 days of delivery..." }) // 2. Switch the sender's agent on. From then on it answers inbound // messages by itself: it searches the knowledge base and replies. await zavu.senders.agent.update(senderId, { enabled: true })
Typical results:
- Repetitive questions answered without your team
- Replies in seconds, 24/7
- When your documents don't cover a question, it says so, as your system prompt tells it to
- Every reply is recorded as an execution you can review
Zavu AI agent technical details.
WhatsApp support best practices
Tone
- Treat customer as a person, not a ticket
- Use emojis sparingly (depends on audience age)
- Short sentences, easy to read on mobile
Response time
- Automatic first reply in < 30 seconds
- Human reply in < 5 min during business hours
- Human reply in < 1h outside hours (announce in away message)
Closing
- Always confirm the question was resolved
- Send satisfaction survey at the end
- Label "Resolved" to close the conversation
Privacy
- Don't ask for sensitive data (SSN, password, card) via WhatsApp
- Use links to secure areas of your site
- Delete conversations per GDPR/CCPA (Zavu has configurable retention)
Common mistakes
1. Mixing support and marketing on the same number. Customer who got a promo on the support number gets angry. Use separate numbers.
2. Slow first response. Configure automatic reply in < 30s, always.
3. Obvious robot. Bad AI annoys. If the answer is "I don't understand, can I help with something else?" repeatedly, transfer to human.
4. Not measuring. Without metrics, you don't know if you're good or bad. Start with average response time.
5. No clear opt-out. Customer has the right to leave the list. Have "Reply STOP to unsubscribe" always.
Comparison: Zavu vs WhatsApp Business app
| App WhatsApp Business | Zavu (API + dashboard) | |
|---|---|---|
| Cost | Free | From $25/month |
| Multiple agents | 1 only | Unlimited |
| Sustainable volume | ~50/day | No limit |
| Metrics | Basic | Complete + real-time |
| CRM integration | Manual | Native |
| AI agent | No | Yes |
| Auto-messages | 2 (greeting + away) | Unlimited |
| Templates | Manual via Business Suite | API + auto-approval |
| Webhook | No | Yes |
Related resources
- How WhatsApp Business works
- WhatsApp Business auto-reply
- WhatsApp Cloud API tutorial
- AI agents for WhatsApp
- Complete documentation
Conclusion
WhatsApp Business support at scale doesn't fit in the free app. When you pass ~50 conversations/day or need multiple agents, the API with multi-agent dashboard is the path. Adding AI with RAG, the agent takes the repetitive questions and frees the team for complex cases. Zavu delivers everything in a single provider — start with the free tier of 2,000 messages, scale up when needed.