WhatsApp customer service: setup, team workflow and AI help

Many customers would rather send a WhatsApp message than call or write an email. They ask about a delivery, a size or an invoice, and expect an answer in the same chat. For a store, that means one phone number receives questions all day, and the person holding the phone becomes the bottleneck. This guide explains what you need to run WhatsApp customer service as a team, how to set up a shared inbox, what a good daily workflow looks like, where an AI agent helps, and which mistakes to avoid. It belongs to the WhatsApp Business guide.

Why customers use WhatsApp for support

WhatsApp is already open on the customer's phone, so there is nothing to install and no ticket form to fill. A few things follow from that:

• Short questions: "Where is my order?" or "Does this come in blue?" arrive as a single line.
• Context stays in one thread: the customer scrolls up to find the invoice you sent last month.
• Photos and voice notes: a customer can show a damaged item instead of describing it.

The same habits create pressure. A message that waits a day feels ignored, so the way you organise replies matters more than the channel itself.

What you need before you start

Three things must be ready:

• A WhatsApp Business API connection: the free app suits one person with one phone. A team needs the API, connected through Meta's signup flow. See what WhatsApp Business is and the costs in opening an account and pricing.
• A shared inbox: software where the whole team reads and answers the same chats, each with their own login.
• Written answers: shipping, returns, sizing and opening hours, in words you are happy for a customer to read.

Set up a shared inbox

In StoreChart, the WhatsApp inbox connects through Meta's signup flow from inside the product. After that:

• Everyone replies from one number: any user with access to Communication can read and answer, without sharing a phone or a password.
• The customer's context sits beside the chat: the phone number is matched to orders and total spend, or to an open lead.
• Unread chats are flagged: each chat shows how many new messages it has.
• Search: by customer name, phone number or text from the latest message.
• Old chats: if you agree when connecting, Meta sends up to 180 days of history from the Business app.

Note what the inbox does not do: there is no per-agent assignment. Two people can answer the same chat, so agree on a split in advance, for example by shift or by language.

A daily workflow that works

A simple routine keeps replies fast without a contact-centre budget:

1. Start with unread chats. Sort the morning backlog into quick answers, orders to check, and requests that need a decision.
2. Answer the repeated questions with saved replies. Type a slash and the shortcut, edit the text, send.
3. Check the order beside the chat before you answer a delivery question.
4. Leave a clear next step when you cannot solve it at once: who will check, and when the customer hears back.
5. Mark chats you still owe a reply as unread, so they stand out again.
6. Review the 24-hour window on older chats. If it has closed, reopen with an approved template.

Example: a Monday morning at a sports store

The store starts the day with 30 unread chats. Twelve are sizing questions, closed with a saved reply and a size chart. Eight are "where is my order", answered from the orders shown beside the chat. The last ten are mixed requests that take most of the team's time. By evening, only chats waiting on the warehouse stay unread, and they come first tomorrow.

The figures in this article's examples are illustrative only and do not reflect actual customer data.

Where an AI agent helps

Most of those 30 chats were the same few questions. An AI agent can take them:

• Repeated policy questions: shipping times, return rules, opening hours.
• After-hours messages: it answers while the team is away, within working hours you set.
• Drafts for your team: a button in the composer drafts a reply from your business knowledge, and the agent decides nothing until you send it.

The StoreChart WhatsApp AI agent runs in the same inbox. In supervised mode it replies until someone from your team writes in that chat, then stops there. It does not read orders or live stock, so those questions go to your team. For the wider picture, see the guide to AI customer service.

Reopening a conversation after 24 hours

Sooner or later a customer goes quiet and you need to follow up, or you finish checking an order the next day. When the window has closed, free text is blocked. StoreChart's inbox says so in the composer and offers to reopen the conversation with an approved template. Prepare one service template in advance, short and neutral, such as "Hi, we have an update on your question. Reply here and we will continue." Once the customer answers, a new 24-hour window opens and you can write freely again.

Measure what matters

You do not need a dashboard to learn from WhatsApp support. A few habits are enough:

• First reply time: note how long the oldest unread chat waited at the start of each shift.
• Unread at the start of a shift: a number that keeps growing means the team is short or the questions are too repetitive.
• Repeated questions: write down what you answered three times this week. Each one deserves a saved reply, or an entry in the agent's knowledge.
• Wrong answers: every correction a customer sends back is a fix to make in your saved replies and policies.

Read ten real conversations each week. Most improvements come from there, not from a report.

Rules to respect

• The 24-hour window: free text only within 24 hours of the customer's last message.
• Templates and opt-in: to write first, you need an approved template and the customer's agreement.
• Permissions: not everyone needs to read customer chats. Set inbox access in team permissions.

Common mistakes

• One phone for everyone: the first person to leave takes the conversations with them.
• No split of work: two people answer the same question differently.
• Answering from memory: a wrong delivery promise binds the business. Write policies down once and reuse them.
• Letting the unread count grow: a number nobody checks teaches customers not to write.
• Automating money decisions: refunds and compensation stay with a person.