WhatsApp Auto-Reply Setup: What to Automate and What to Skip

A furniture showroom we know turned on an auto-reply that said 'Thanks for contacting us! Our team will get back to you soon.' It went out to every message, including 'I need to change my delivery address, it's arriving tomorrow.' The customer read the cheerful reply, assumed someone had seen the problem, and waited. Nobody had.

That is the usual failure. Auto-replies are easy to switch on and hard to get right, because the question is not 'can we reply automatically' but 'which messages deserve a machine's answer'. This guide is how we think about that when we set things up on WA Getway.

Start by sorting your incoming messages

Before touching any setting, scroll through a week of your real chats and sort them into three piles. Messages that need a fixed answer (address, timings, price list, order status). Messages that need a person (complaints, custom quotes, anything with money in dispute). Messages that need nothing (thanks, ok, thumbs up).

Automation belongs on the first pile only. The second pile should reach a human quickly, and the third pile should be left alone. A reply to 'thanks' with another 'thanks' is how you end up in a loop with someone else's bot.

The auto-replies worth having

• First-contact acknowledgement for new numbers: who you are, what you can help with, and the one or two questions that move things forward (name, city, what they are looking for).
• Outside-hours notice that states when someone will actually respond. A specific time beats 'soon'. We cover this separately in business hours and out-of-office replies.
• Keyword answers for the five questions you answer daily: location, timings, pricing sheet, how to book, how to pay.
• Confirmation of receipt after a customer sends a form, a photo or a document, so they know it arrived.

The ones we would skip

Skip replies to every single message in a conversation. One acknowledgement per conversation is polite; one per message is spam you sent to yourself. Skip replies that pretend to be a person ('Hi, this is Priya!') when it is not. Customers forgive a bot that admits it is one; they do not forgive finding out later.

Also skip long menus. A reply that says 'Press 1 for sales, 2 for support, 3 for billing...' works on a phone line because you cannot scroll a phone call. On WhatsApp, a customer would rather type 'invoice copy' and get it. Keep the reply to three options at most, and always include a way to reach a human.

Write the replies like you talk

Read your draft aloud. If you would never say it to someone at your counter, rewrite it. 'We have received your query and will revert at the earliest' is the written equivalent of a recorded phone voice. Try: 'Got your message. We open at 10 tomorrow and someone will reply by 11.'

Put the useful information first. Customers read the first line in the notification preview, and many never open the chat. If your timings are the answer, lead with the timings.

Test before it faces a customer

1. 1Send messages from a personal number to your own business number and check each rule fires once, not twice.
2. 2Try odd inputs: a single emoji, a voice note, an image, a message in another language. Decide what happens in each case, even if the answer is 'nothing'.
3. 3Check that your rules do not fight each other. A keyword rule and a general greeting both matching the same message is the most common bug we see.
4. 4Ask a colleague who did not build it to try to break it.

Our walkthrough on testing a WhatsApp integration without spamming customers goes deeper on keeping tests away from real contacts.

Know the limits of the channel

Two things to check with whichever tool you use. First, what kinds of rules it supports: plain keyword match, any-message, time-based. Not every tool offers all of them, so confirm before you design around one. Second, what the platform allows. If you use the official Meta Cloud API, free-form replies are tied to the customer-service window, and anything outside it needs an approved template. See the 24-hour customer service window post, and check Meta's current documentation for the rules as they stand today.

If you run your numbers through a QR-based session instead, the rules are looser but so is the safety net. Heavy automated replies at volume can attract spam reports, so keep automation modest and useful.

Review it monthly

Auto-replies rot. Your timings change, a price list goes out of date, a festival closure is not reflected. Put a monthly reminder in your calendar to read every automated message as if you were a new customer. Delete any that nobody triggered.

Automation should take the repetitive questions off your plate so your team has time for the conversations that earn money. If you want help setting rules around your own workflow, or something custom through our API, get in touch and we will talk it through.