WhatsApp bot for online stores: how to answer and close orders around the clock
How to use an AI WhatsApp bot in an online store: reply at any hour, quote prices, collect shipping details and hand the sale to your team at the right time.

In this article
It's 11:40 p.m. Someone sees your ad on Instagram, likes the product and sends a message: "Price? Do you ship to my city?"
Your team went home hours ago. The message sits unread until 9:00 a.m.
By then, that person has already bought from another store. The one that answered first.
For an online store, the chat is the checkout counter. And a counter that closes at 6 p.m. loses every sale that walks in after.
In this article you'll see why online stores lose sales on WhatsApp, what an AI bot actually does at each step of an order, and which numbers tell you whether it's working.
In short: an AI WhatsApp bot for online stores replies within seconds at any hour, understands what each customer wants (even from photos or voice notes), quotes with your rules, collects shipping details and hands the sale to your team when a human needs to close. It works best when you write down your catalog and conditions as if you were training a new salesperson.
Why online stores lose sales in chat
People who buy through chat aren't patient. They message when they see the ad, not when your team is available.
And messaging a business is now a mass habit. In 2022, Meta said on its developer blog that more than 1 billion users connect with a business account every week across WhatsApp, Messenger and Instagram Direct. It's a global figure, and it comes from Meta itself, but it shows that talking to a business over chat is now an everyday thing.
At Optimus, an odor-eliminator brand in Chile that sells on Messenger, WhatsApp and Instagram, 61% of replies went out outside business hours. Most of the conversation happened at night, before dawn or on weekends.
If those conversations wait until the next morning, they go cold. A study published in Harvard Business Review in 2011 found that most companies don't respond to their leads nearly fast enough; we walk through its numbers and its limits in why answering WhatsApp leads fast is worth money.
There are two more problems:
- The questions repeat. Price, sizes, shipping, payment methods, stock. Your team answers the same thing a hundred times a day.
- Customers don't write like a form. They send a photo of the product they saw, a two-minute voice note, or "what about for a small apartment?". A button bot understands none of that.
What an AI WhatsApp bot does for a store
A menu bot makes people pick "1, 2 or 3". An AI agent holds a conversation. We break down the difference in AI agent vs. button chatbot.
In an online store, a well-configured agent:
- Replies instantly, day and night, on WhatsApp, Instagram and Messenger.
- Understands photos, voice notes and documents: it recognizes a product from a picture or gets the gist of a voice note.
- Recommends the right option based on what the customer needs.
- Quotes with your rules: by quantity, by bundle, by shipping zone.
- Collects order details: name, address, city and email for the receipt or invoice.
- Hands the sale to a person when there's a negotiation, a special charge or a complaint.
- Logs everything in your CRM, so your team knows who wrote, what they asked for and where it stands.
Think of it as a salesperson who never gets tired, knows your catalog by heart and knows when to call the manager.
How an order moves through chat, step by step
1. Reply within seconds
The first reply matters most. If it's slow, the customer keeps browsing other stores.
At Optimus, the agent's typical response time was 70 seconds, at any hour. And that number includes a deliberate short pause to read the full message before replying.
What to do: measure today how long your store takes to answer a first message after hours. If the answer is "the next morning," that's your first leak.
2. Understand what the customer wants
"Price?" isn't a complete question. The price depends on which product, how many units and where it ships.
A good agent asks just enough to recommend well, without grilling anyone: one or two questions, in plain language.
What to do: write down the three questions your best salesperson always asks before quoting. Those are the ones the agent should ask.
3. Quote the right price, with your rules
This is where generic bots fail most: they invent discounts or mix old prices with new ones.
The fix is written rules: your current price list, bundles, shipping costs by zone and what the agent must not offer.
What to do: build a table with your current prices and conditions, plus a list of offers that have ended. The agent needs both.
4. Collect shipping details
When the customer says "I'll take it," the agent asks for the details in order: name, address, city, contact phone. No external forms, no lost conversation.
Every extra field counts. Baymard Institute, which researches e-commerce usability, estimates that an ideal checkout can ask for as few as 7 or 8 form fields, while the average US checkout shows 14.88 (page updated in 2025). It measures web forms, not chats, but the idea carries over: don't ask for anything you don't need to ship.
What to do: decide which details you need to ship and in what order you ask for them. Fewer fields, fewer drop-offs.
5. Hand the sale to your team at the right moment
Some things shouldn't be automated: off-list discounts, large orders, upset customers. The agent spots them and passes the chat to a person, with full context so nobody has to ask again.
What to do: list the situations where you want a human to take over. That list matters as much as the catalog.
6. Log every step in the CRM
If the order lives only in the chat, it gets lost. When the agent moves each conversation through your CRM stages (for example, in GoHighLevel), you can see how many customers asked, how many got a price and how many bought.
What to do: create your chat-sales stages in your CRM. No stages, no way to measure.
Which numbers to watch
A bot that "chats nicely" is useless if it doesn't move sales forward. That's why you should look at the whole funnel, not just message counts.
Here's the Optimus funnel over 61 days, as recorded in its CRM:
| Stage | Customers |
|---|---|
| Served by the agent | 2,385 |
| Pre-qualified by the agent | 1,076 |
| Received a price | 231 |
| Handed off to the sales team | 222 |
| Left their shipping details | 149 |
How to read it:
- Served to pre-qualified tells you whether the agent understands what people want. A big drop means you should review its questions.
- Pre-qualified to price tells you whether the agent gets to the offer. If it drops, it may be asking too much before quoting.
- Handoffs and shipping details are the bridge to closing. That's where your team needs to be on it.
An honest note: these stages aren't closed sales. They're steps measured in the CRM. The final sale also depends on price, stock and how quickly your team picks up what the agent leaves ready.
When a WhatsApp bot is NOT a good fit
Not every store needs one. It's probably not worth it if:
- You get few messages a day and your team answers them quickly, even after hours.
- Every sale needs long technical advice or an in-person visit before quoting.
- Your prices change daily and there's no list you can write down.
In those cases, sort out the process first. Automating a messy process just makes the mess faster.
How to start this week
- Measure how many messages arrive after hours and how long you take to answer them.
- Write down your catalog with current prices, shipping and payment methods.
- Note the questions your best salesperson asks and the situations where a person should step in.
- Create your sales stages in the CRM.
- Test the agent with real conversations before switching it on for everyone.
For the full process, see how to automate WhatsApp with AI.
Frequently asked questions
Can a WhatsApp bot close sales on its own?
It can carry the conversation all the way: recommend, quote and collect the order details. What you should decide is which part the agent closes and which part a person closes, such as special discounts or large orders.
Does it also work on Instagram and Messenger?
Yes. Online-store customers write on whichever channel showed them the ad. At Optimus, most messages came through Messenger, followed by WhatsApp and Instagram, and the same agent handled all three.
What if the customer sends a photo or a voice note?
An AI agent can handle both: recognize the product in a photo and understand what the customer says in a voice note, then reply as if it had read a text message.
How long does it take to get it running?
It depends on how clear your catalog and sales rules are. With Merkhify, setup takes 7 days: that includes loading your products, prices and conditions, connecting your channels and testing with real conversations.
Your store is already open 24 hours a day. The question is whether anyone is at the counter when the customer walks in.


