Conversational Commerce in Practice: The Four Pieces a Chat Sale Needs

Most businesses already sell in conversations. A customer messages on Instagram, asks
about a product, and the business answers. Somewhere in the next few messages the business
sends a link to its website, and the conversation stops being a conversation.

Conversational commerce is the version where it does not stop. The customer chooses,
pays and gets delivery updates in the same thread they started in. That is the whole
definition. The money moves inside the conversation.

This piece is about what has to be true for that to work. Not the idea of it. The four
specific parts, what each one does, and what you will see go wrong when one of them is
missing.

Why the website step costs you orders

Every handoff is an exit. A customer who is three messages into a conversation with you
has momentum. Send them to a product page and they have to find the item again, pick the
variant again, create an account, re-enter an address and face a delivery charge they have
not seen yet. Some of them finish. Most do not.

The Baymard Institute keeps a running average of published cart abandonment studies. As of its
September 2025 update the documented average was 70.22%. That figure is a mean across
many studies and store types, not a benchmark for your shop, and it includes shoppers who
were only ever price-checking. But the direction is not in dispute. The checkout is where
intent goes to die.

A sale that completes in the thread never reaches that checkout. There is no cart to
abandon.

Piece one: something to choose from

The conversation needs to be able to read a catalogue. Not a PDF, not a screenshot of a
price list. Something structured, with names, prices, variants and stock.

There are three reasonable versions of this, in increasing order of effort.

• A Google Sheet. One tab for products, one for orders. Good enough to
  start selling, and the cheapest route for a seller with no storefront at all. It is not an
  order management system and it will not stay sufficient forever. It is the right first
  step for someone whose current catalogue is photos in a WhatsApp album.
• A Meta or Instagram catalogue. If you already sell on Instagram, this
  is where your products are. It is attached to the channel the customer is already in.
• A real store. Shopify or WordPress. Now the conversation can read
  live stock, real variant prices and your existing shipping rates, and the order it creates
  lands in the system you already run your business from.

The common failure here is subtler than having no catalogue. It is having a catalogue
the conversation cannot see. A tool that answers from your website’s text can tell a
customer what a product is. It cannot tell them whether the medium is in stock, because
that is not written on the page. See retrieval-augmented generation for why those are different problems.

Piece two: a way to pay

The customer has chosen. Now they need to pay without leaving. That is a payment link: a
single-use checkout sent into the thread.

Three things matter about how that is set up, and they are worth asking any vendor
about directly.

Who holds the money. Settlement should land in your own account with
your own payment provider. AeroChat never sees card numbers and never holds your funds.
Money settles directly into your Airwallex balance.

How payment is confirmed. You cannot run fulfilment off the customer
saying “paid already”. Confirmation has to reach your systems. AeroChat confirms every
payment twice, once the moment the Airwallex webhook fires, and again when the customer
lands on the confirmation page.

What happens next. A paid link that does not create an order or tell
anyone to pack something has moved your admin, not removed it.

Payment links work on the channels customers actually use: your website chat, WhatsApp,
Messenger, Instagram, Telegram and email.

Piece three: a way to get it there

This is the part most discussions of conversational commerce skip, and it is the part
that decides whether the customer ever messages you again.

They have paid inside a chat. The natural next question arrives in the same thread
within a day: when is it coming. If your answer requires someone to open a courier
dashboard, copy a tracking number and paste it back, you have automated the sale and left
the aftermath manual. Order-status questions are the most repetitive thing most sellers
deal with, and they arrive in volume after exactly this kind of sale.

So delivery needs to be part of the same system. In AeroChat that means booking
delivery with POS Malaysia directly: creating a shipment and label, booking, rescheduling
or cancelling a pickup, tracking a parcel, or quoting an international rate. Delivery
updates then go back to the shopper on the channel they used. AeroChat connects to the
payment provider and to the courier separately. Neither one goes through the other.

Piece four: the conversation that runs it

Three pieces of infrastructure do nothing on their own. Something has to read what the
customer wrote, work out what they want, check the catalogue, send the right link,
register the payment, book the delivery and answer the follow-up.

This is the piece businesses are usually missing, and it is the reason the other three
sit idle. You can have a Shopify store, a payment provider and a courier account, and
still be copying prices into Instagram by hand at 11pm.

What it has to be able to do, concretely:

1. Understand a message written the way customers actually write. Short, unpunctuated,
   often mixing two languages in one sentence.
2. Read your live data, not your marketing copy.
3. Take an action in a real system, rather than describing one.
4. Stop at the limits you set. Refund ceilings, discount rules, delivery promises.
5. Hand over to a person when it should, with the conversation intact.

In AeroChat you set this up by describing the workflow in plain English. There is no
flow builder to learn. The limits you set on refunds, discounts and promises are applied
by the system itself rather than left to the model’s judgment.

How to tell which piece you are short of

Take twenty recent conversations that did not end in an order. For each one, find the
last message before the customer went quiet, and mark which piece failed.

In our experience the last row is the biggest pile, and the one people least expect.
Roughly half of consumers expect a reply within an hour (Twilio consumer preferences report, which sells messaging
software, so treat it as vendor data). The orders you are losing are mostly not lost to a
competitor’s better product. They are lost to the gap between 7pm and 9am.

Is any of this actually normal yet?

In Southeast Asia, buying in a chat was normal before the current round of AI tooling
existed. A Facebook and Boston Consulting Group study published in 2019
found the share of online shoppers who had bought through a chat conversation was 40% in
Thailand, 36% in Vietnam, 29% in Indonesia, 26% in Malaysia and 23% in the Philippines
(reported via KrASIA). That study is old and should always be dated, and no recent public figure isolates
chat purchases by country in the region.

On preference, a Kantar survey of 11,056 adults across 22 markets commissioned by Meta,
with fieldwork from April to September 2025, found 73.3% of consumers prefer messaging
when contacting a business and 72.4% are more likely to buy from a brand that offers it
(Meta, State of Business Messaging). Meta sells business messaging, so that is the seller’s own research.

On size, Juniper Research, October 2023 put global retail spend through conversational commerce channels at
US$11.4 billion in 2023, forecast to reach US$43 billion by 2028. Retail spend only. You
will see a much larger “US$290 billion by 2025” figure circulating. That was Juniper’s
2021 forecast, it covered a far wider scope, and it has been superseded. Do not use
it.

What to do with this

Do not start by buying software. Start by running the twenty-conversation check above.
It takes an hour and it tells you which of the four pieces to fix first. Most businesses
discover they already own three of them.

Read next

• Conversational commerce (definition)
• Payment link
• Cart abandonment
• Retrieval-augmented generation (RAG)
• Payment links, catalogues and delivery booking, in detail
• How a small seller starts, with a Google Sheet

Start for free

AeroChat runs customer service and sales workflows you describe in plain English. There is a free plan, so there is nothing to decide today.

Start for free on the Shopify App Store

Not on Shopify? Create an AeroChat account here.