---
title: "How to Design a Multi-Agent WhatsApp Customer Support Workflow"
description: "Design a multi-agent WhatsApp customer support workflow: routing, escalation, the 24-hour service window, saved replies, SLAs and Arabic/English teams."
canonical: "https://omnichannels.ai/blog/whatsapp-customer-support-workflow/"
language: "en"
updated: "2026-10-10"
publisher: "BroadNet Technologies"
brand: "OmniChannels.ai"
category: "support"
primaryQuery: "whatsapp customer support workflow"
keywords: ["whatsapp support team", "whatsapp business api multiple agents", "whatsapp 24 hour window", "whatsapp support sla", "whatsapp escalation process", "arabic english customer support"]
published: "2026-10-10"
author: "OmniChannels.ai Editorial Team"
reviewer: "BroadNet Messaging Solutions Team"
readingMinutes: 8
answer: "A multi-agent WhatsApp support workflow connects one WhatsApp Business Platform number to a shared inbox, then defines five things: how new chats are routed (by language, topic or customer), who owns each chat, when and how it escalates, how agents use saved replies inside the 24-hour customer service window and approved templates outside it, and which response and resolution targets the team is measured against."
takeaways: ["Multi-agent WhatsApp support requires the WhatsApp Business Platform (API) connected to a shared inbox; the WhatsApp Business App is not designed for team-scale routing.", "Free-form replies are allowed only within 24 hours of the customer's last message; after that the business must use a Meta-approved message template.", "Route Arabic and English chats to separate queues from the first message, and keep bilingual agents as overflow rather than the default.", "Escalations should change the owner or add a named specialist with a note, never silently forward the chat.", "Set SLAs per priority and channel and measure them with medians from the first customer message, including out-of-hours time if you promise 24/7."]
faq: 
  - q: "Can several agents answer the same WhatsApp business number?"
    a: "Yes, with the WhatsApp Business Platform (API). The number connects to an inbox or helpdesk tool and many agents can work its conversations in parallel, each chat assigned to one owner. The WhatsApp Business App is aimed at small businesses with a limited number of linked devices and lacks queue routing, so support teams that grow beyond a few people usually move to the API."
  - q: "What should agents do when the 24-hour window has expired?"
    a: "They cannot send a free-form message. They send a pre-approved template, usually a utility template such as an update on the customer's case, inviting the customer to reply. When the customer replies, a new 24-hour window opens and normal conversation resumes. Templates are reviewed by Meta and may be charged by category and recipient market, so keep a small approved set ready."
  - q: "How do you route WhatsApp chats by language?"
    a: "Detect the language of the first message, or offer an Arabic or English choice through a template or quick-reply button, then route to the matching queue. Keep the customer's language on the contact record so later conversations go straight to the right team. Bilingual agents act as overflow when one queue is busy. Avoid asking customers to switch language mid-conversation."
  - q: "Is WhatsApp customer support free?"
    a: "Not entirely. Meta charges per delivered template message by category and market. According to Meta's pricing documentation, from 1 October 2026 Meta charges service messages after the first 1,000 free per business phone number per month, and charges utility templates sent inside the customer service window. Confirm current terms in Meta's official pricing documentation and rate card before budgeting."
  - q: "What is a good first response time on WhatsApp?"
    a: "There is no official benchmark. Customers treat WhatsApp as a chat channel, so teams typically aim for minutes during staffed hours rather than the hours common on email. Set your own target from your staffing, publish your hours, and acknowledge out-of-hours messages automatically with expectations so the measured response time reflects reality."
sources: 
  - title: "Send messages and customer service windows — WhatsApp Cloud API, Meta for Developers"
    url: "https://developers.facebook.com/docs/whatsapp/cloud-api/guides/send-messages"
  - title: "Message templates — WhatsApp Business Platform, Meta for Developers"
    url: "https://developers.facebook.com/docs/whatsapp/business-management-api/message-templates"
  - title: "WhatsApp Business Platform pricing — Meta for Developers"
    url: "https://developers.facebook.com/documentation/business-messaging/whatsapp/pricing"
---

# How to Design a Multi-Agent WhatsApp Customer Support Workflow

Canonical: https://omnichannels.ai/blog/whatsapp-customer-support-workflow/
Language: en
Publisher: OmniChannels.ai by BroadNet Technologies

This guide is for support leaders who need several agents, often across Arabic and English shifts, to answer customers on one WhatsApp number without chaos. It covers the workflow decisions in the order you should make them: routing, ownership, escalation, the service window, saved replies, and service level agreements (SLAs).

## What do you need before multiple agents can answer WhatsApp?

You need the WhatsApp Business Platform (the API), a verified business phone number, and a shared inbox or helpdesk connected to it. The WhatsApp Business App is built for small teams on a handful of linked devices and has no queue routing or assignment reporting.

| Need | WhatsApp Business App | WhatsApp Business Platform + inbox |
|---|---|---|
| Many agents in parallel | Limited linked devices | Yes, limited by inbox tool |
| Assignment and queues | No | Yes |
| Templates outside 24h | Not applicable in the same way | Meta-approved templates |
| Reporting per agent | No | Yes, in the inbox tool |
| Integration with CRM | Manual | Via APIs and webhooks |

## How should new WhatsApp chats be routed?

Route on the first message using the cheapest reliable signal: entry point, language, keyword or known customer. Each signal decides a queue; a queue decides the eligible agents.

### Routing signals in order of reliability

1. **Known customer.** If the number matches a CRM contact with an account owner or an open case, route to that owner or case team.
2. **Entry point.** A WhatsApp link on the order confirmation page or a QR code in the showroom tells you intent before the customer types.
3. **Language.** Detect Arabic versus English from the first message or ask with a quick-reply choice.
4. **Keyword or menu choice.** Order number pattern, "refund", "invoice", or a menu selection.
5. **Fallback.** Anything unclear goes to a general triage queue with a short target.

Inside a queue, assign by load (fewest open chats among online agents) rather than strict round robin, so a slow complex case does not stall new arrivals.

## How should escalation work?

Escalation moves a chat to someone with more authority or expertise, with context, and with the clock still running. Define it by trigger, destination and handover format.

| Trigger | Escalate to | Handover rule |
|---|---|---|
| Refund or credit above agent limit | Team lead | Owner stays agent; lead adds approval note |
| Technical fault needing investigation | Tier 2 queue | Reassign with summary note; agent tells customer who is taking over |
| Complaint mentioning legal action or regulator | Supervisor | Reassign immediately; no further replies from tier 1 |
| VIP or key account | Account manager | Reassign; customer is told by name |
| Abusive messages | Supervisor | Supervisor decides on blocking per policy |

Two rules prevent most escalation failures: the customer is always told what happens next, and the internal note answers "what has been tried, what is needed, by when".

## How do you work with the 24-hour customer service window?

When a customer messages you, a 24-hour customer service window opens in which you can send free-form messages. After it closes, you can only reach the customer with a Meta-approved message template until they reply again.

Design the workflow around this clock:

- **Show the remaining window** on each conversation and sort the queue so chats close to expiry rise to the top.
- **Resolve or set expectations inside the window.** If a case needs two days, tell the customer now and agree that you will follow up with an update.
- **Prepare utility templates** for predictable follow-ups: "Your case {{1}} has been updated, reply to continue", delivery rescheduled, document received.
- **Do not use marketing templates for service follow-ups.** Categories are set by Meta's rules; misuse risks rejection or recategorisation.

On cost: Meta charges per delivered template message by category and recipient market. According to Meta's pricing documentation, from 1 October 2026 Meta charges service messages after the first 1,000 free per business phone number per month, and charges utility templates sent inside the customer service window. Confirm in Meta's current rate card before you model costs.

## How should saved replies be organised?

Saved replies are internal shortcuts for answers agents send often; they are not templates and need no Meta approval, but they can only be sent inside the window.

A practical structure:

- **Name by intent and language**: `delivery-time-ar`, `delivery-time-en`.
- **Keep variables obvious**: `[order number]`, `[date]`, so agents never send a placeholder.
- **Limit to the top 20 to 40 intents**; a library of 300 is never used correctly.
- **Write natively in each language.** Gulf customers notice translated Arabic; have a native writer draft the Arabic set rather than translating the English set word for word.
- **Review monthly** against the questions agents actually receive.

## How do you set SLAs for WhatsApp support?

Set separate targets for first response and resolution, per priority, measured in business hours or calendar hours depending on what you promise.

| Priority | Example definition | First response target | Resolution target |
|---|---|---|---|
| P1 | Service down, payment failed, safety | Set in minutes | Same day |
| P2 | Order problem, account access | Set in minutes to an hour | 1 business day |
| P3 | General question, pre-sales | Within staffed hours | 2 business days |

The numbers are placeholders for your own decisions; there is no official WhatsApp benchmark. Measure with medians and the 90th percentile, start the clock at the customer's first message, and exclude nothing silently: if out-of-hours chats are excluded, report that.

## How do you run Arabic and English teams on one number?

Use one number and separate queues, not separate numbers. Customers should not have to know which number serves which language.

1. Capture language on the first message and store it on the contact.
2. Staff each queue to its own volume curve; Arabic and English demand often peak at different times.
3. Make bilingual agents overflow for both queues during peaks.
4. Maintain parallel saved reply and template sets, each approved in its language.
5. Report KPIs per language queue to spot an under-staffed team early.
6. Align weekend and holiday hours with your market; Gulf workweeks and religious holidays affect volume.

## Worked example: a Riyadh telecom reseller

"Sahab Connect", a synthetic reseller of business internet in Saudi Arabia, receives WhatsApp support messages from SMEs. Twelve agents work two shifts; eight are Arabic-first and four English-first.

- A customer sends "الانترنت مقطوع" with an account number. The workflow matches the number to a CRM account, detects Arabic, and routes to the Arabic tier 1 queue, where the agent with the fewest open chats receives it.
- The agent runs a saved troubleshooting checklist; the fault persists. She escalates to tier 2 with a note covering line test results and the customer's availability, and tells the customer an engineer will reply within the hour.
- The engineer needs a site visit the next day. Before the window closes, he confirms the slot. The next morning, outside the window, the dispatcher sends an approved utility template with the technician's arrival time; the customer replies, which reopens the window.

## Common mistakes to avoid

- **Treating the API like a personal phone.** Without ownership rules, agents reply over each other.
- **Escalating by forwarding screenshots** to a manager's personal WhatsApp, which breaks history and privacy controls.
- **Letting windows expire silently**, then spending templates to restart conversations that could have been closed.
- **One SLA for everything**, which hides P1 failures in a healthy average.
- **Templates written as sales messages** when the purpose is service, risking rejection.
- **No out-of-hours message**, so customers wait in silence overnight.

## How OmniChannels.ai helps

OmniChannels.ai, powered by BroadNet Technologies, connects your number to the [WhatsApp Business API](/channels/whatsapp-business-api/) through Meta Embedded Signup and brings conversations into the [Omni Inbox](/platform/unified-inbox/) alongside Instagram, Telegram and Email, with agent assignment and tagging. Agents reply with media, attachments and approved templates, and [AI agents](/platform/ai-agents/) can answer routine questions and hand over to a human. The platform supports Arabic and English, delivery reports with CSV export help you review template follow-ups, and CRM routing uses the public [REST API and webhooks](/developers/).