Automate customer support on WhatsApp

Start with a shared inbox

Support automation only works if the people behind it are organised. A shared inbox puts every WhatsApp chat in one place, where each teammate signs in with their own account. Chats can be assigned to a person, labelled by topic, snoozed until later and resolved when done. Private notes let colleagues add context the customer does not see.

In WaKite, Instagram DMs sit in the same inbox, so one team can cover both channels. Shortcuts help agents answer common questions by hand: type a slash and the shortcut name to insert a saved text, a file, a link or a contact card. Bulk actions let a supervisor assign or resolve many chats at once after a busy day.

Route chats with a menu

The first automated step in support is usually a menu that asks what the customer needs. A list menu can offer up to ten topics, such as "Order problem", "Billing", "Technical help" and "Something else". Each answer can lead to a short self-help path, or straight to an assign step that gives the chat to the right person, with a label for the topic.

Keep the menu honest. If a topic always ends with a person, send it there at once instead of making the customer tap through three screens first. Add a "Talk to a person" option on every path, and remember that customers can also just type. A fallback rule catches messages that match nothing and offers the menu again.

• Up to ten topics in one list menu
• Assign by topic, with a label
• A person on every path
• A fallback for anything unexpected

Away messages and business hours

Set your business hours for each day of the week, and an away trigger answers messages that arrive outside them. A good away message says when you will reply, offers the self-help options that work without a person, and can collect the details your team will need, so the morning shift starts with complete requests rather than a pile of "hi".

Do not promise what you cannot keep. If the team replies within two hours on weekdays, say that, and say what happens on Sundays. Customers are patient when they know what to expect. When a teammate replies, automation steps back for that chat so the bot and the person do not talk over each other.

AI answers from your knowledge base

Many support questions are already answered in your help pages, policies and product manuals. WaKite lets you build a knowledge base of articles, typed in or uploaded as text, Markdown, CSV, HTML, Word or PDF files. The AI step in a flow searches it and writes a reply from what it finds, in the tone you choose, from friendly to brief.

Keep the AI on a short lead. Let it answer only from your knowledge base, and connect its hand-off path so that, when it is not sure, the chat goes to a person. The step can tell customers they are chatting with an automated assistant and that typing "agent" reaches a human. The same step can also sort messages by intent or pick out a detail such as an order number.

Reply targets and ratings

A reply-time target, often called an SLA, sets how long a customer may wait for a person. In WaKite you set the target in minutes for the workspace. When a waiting chat goes past it, the team is told the reply is overdue, and a webhook can alert your own systems too. Targets keep busy days from turning into forgotten chats.

After a chat is resolved, a satisfaction survey can ask the customer to rate it with three buttons: great, okay or poor. You can change the question. It is sent only while the 24-hour window is open, and not more than once a day per chat. The rating is saved with the chat and the agent who handled it, so you can thank people and spot problems.

Improve one step at a time

Each week, read a sample of chats the bot handed over and chats that were rated poor. Look for patterns: a question the knowledge base cannot answer, a menu option nobody understands, a topic that always goes to the wrong person. Fix the single biggest problem first, publish, and check again the following week.

For long or messy conversations, an AI summary of the chat helps whoever picks it up next, and it is made only when someone presses the button. Add new knowledge base articles when the same question reaches people twice. Over time the bot handles more of the routine work, and your team spends its time on the conversations that really need care.

Step by step

1. 1
   Invite your team

   Add teammates to the workspace so each has their own login to the shared inbox.

2. 2
   Build the support menu

   Create a list menu of topics that assigns and labels each chat.

3. 3
   Set hours and away message

   Enter your business hours and write an away message that says when you will reply.

4. 4
   Add your knowledge base

   Upload help articles and add an AI step that hands over when it is unsure.

5. 5
   Switch on targets and ratings

   Set a reply-time target and turn on the satisfaction survey for resolved chats.

Questions