Trading & Distribution

WhatsApp Orders into SQL Account or AutoCount: How It Works

How a WhatsApp order becomes a draft sales order in SQL Account or AutoCount: the vendors' documented import routes, the checks in between and the limits.

  • By
  • 7 min read
  • Facts checked
Short answer

A WhatsApp order reaches SQL Account or AutoCount in four steps: the message arrives through the WhatsApp Business Platform, an AI step turns it into order lines, rules check them against your master data, and a draft is written through a route the vendor documents (API, import file or SDK). A person approves the draft before it becomes an invoice.

On this page
  1. How does a WhatsApp order become a draft sales order?
  2. Which import routes does SQL Account document?
  3. Which import routes does AutoCount document?
  4. Which orders should stop for a person?
  5. What does WhatsApp itself allow?
  6. Is there a public example that it pays off?
  7. What should you do before you build?
  8. How we help
  9. Sources
  10. Frequently asked questions

Most distributors in Malaysia take orders the same way: a customer sends "10 ctn Milo 1kg, 5 ctn Nescafe, deliver Thursday" on WhatsApp, and someone in the office types it into SQL Account or AutoCount. The typing is where the day goes, and where the wrong carton size gets shipped.

You can remove the typing without replacing your accounting system. This page shows the flow, the import routes each vendor documents, and the rules that keep bad orders out of your books.

Part of our work for trading and distribution companies. Vendor documentation checked on 16 November 2026. Features depend on your version and licence; confirm with your SQL Account or AutoCount dealer before you build.

How does a WhatsApp order become a draft sales order?

Four steps, and only the last one touches your accounting system.

  1. Receive. On the WhatsApp Business Platform, Meta sends every incoming message to your system as a webhook. The messages webhook reference (Meta, checked November 2026) has separate formats for text, image, audio and document messages, plus an order message when the customer buys from a WhatsApp catalog.
  2. Read. An AI step turns free text, a photo of a handwritten list or a voice note into structured lines: customer, item, quantity, unit, delivery date. Catalog orders arrive structured already and skip this step.
  3. Check. Rules compare each line with your master data: does the customer exist, does the item code exist, is the unit right, is the price the agreed one, is the customer within credit limit.
  4. Write a draft. Clean orders are written as a draft document. A person opens it, checks it and confirms. Anything that failed a rule goes to a short exception list instead.

The design rule we hold to: the AI proposes, your rules decide, a person approves. Nothing posts stock or sends an invoice on its own.

Which import routes does SQL Account document?

SQL Account (eStream Software) documents three ways in.

Route What the documentation says Fits when
SQL Account API REST API with access key and secret key, signed with AWS Signature v4. A Postman collection is provided. The API user can no longer log in to SQL Account directly. You want each order written within seconds
Excel import utility Imports master files and transactions from Excel, CSV and TXT (JSON from version 5.11.10.43). Scheduled auto import for Excel, CSV and TXT from version 5.10.6.32. Batch update; transaction import is free only under 100 rows per file. A batch every 15 or 30 minutes is fast enough
XML import External program for master data and transactions, sales orders included, with a Verify step that lists conflicts before import. You move orders once or twice a day, by hand

Two points from the on-premise setup page decide whether the API route is realistic for you: it needs SQL Account version 5.2025.1038.875 or above, and a public IP address with port forwarding on your router. Many small offices have neither. Then the scheduled Excel or CSV import is the calmer choice, because the file is dropped in a folder and nothing in your network is opened to the internet.

The public pages show invoices in their examples. Check in the Postman collection of your own installation that the sales order endpoint is there before anyone quotes you a price.

Which import routes does AutoCount document?

AutoCount has two different products, and the answer differs.

AutoCount Accounting (desktop). The vendor's wiki states that a sales order has no effects on inventory movement and G/L entries. That makes it the right landing place for a WhatsApp order: nothing moves until staff transfer it to a delivery order or invoice. For writing it, the developer wiki publishes sample code that creates a sales order through AutoCount's own .NET libraries. The page lists assembly versions 1.8 and 1.9, so ask your dealer what applies to your version. This route needs a small Windows service next to your AutoCount server.

AutoCount Cloud Accounting. The Cloud Accounting Integration API works with an API key and a key ID, and each key can be limited to certain methods and tied to one user, so every draft shows who created it. The documentation lists quotations, invoices, credit notes and purchase orders. We found no sales order method in November 2026. In the cloud product the WhatsApp order therefore lands as a draft quotation.

Whichever product you run, never write to the database directly. It skips the validation the vendor built, and your dealer will not support the result.

Which orders should stop for a person?

The exception rules matter more than the AI model. These are the ones we set first.

Rule What happens
Sender's number is not linked to a customer code Held; staff link the number once
Item not matched to exactly one item code Held; staff pick the item, the alias is saved for next time
Unit unclear (carton, box, piece) The system asks the customer on WhatsApp
Quantity far above the customer's usual order Held for a call
Price differs from the customer's price list Held; never corrected automatically
Customer over credit limit or overdue Held for the credit controller
Same order text twice within a short time Flagged as possible duplicate

Aim for a system that drafts the easy orders and is honest about the rest. In the first weeks many orders will stop, because customers write "the usual" and your item names are not theirs. Each exception teaches the alias list, and the share falls. Measure it weekly.

What does WhatsApp itself allow?

Replies such as "Got it, 10 cartons Milo 1kg, delivery Thursday, correct?" are non-template messages. Meta's pricing page (checked November 2026) says they can only be sent inside the 24-hour customer service window that opens with each customer message. Since 1 October 2026 Meta charges for service messages again, with 1,000 free per business phone number each month. A distributor confirming a few dozen orders a day will feel that in the budget only once volumes grow.

Customer names, phone numbers and addresses in those chats are personal data. Store them in your own system, limit who sees them and read our PDPA and AI guide before you send them to an AI service abroad.

Is there a public example that it pays off?

Coca-Cola Brazil is the best documented one. According to an iTnews feature written by the builder, CI&T (October 2020, checked November 2026), a WhatsApp ordering bot for small retailers was built in 20 days in 2019. Retailers using it ordered on average three more product types per order, and Coca-Cola's revenue from those retailers rose by 25 percent. The figures come from the vendor, and a bottler is not a 20-person distributor. The mechanism carries over: customers already live on WhatsApp, so the order arrives faster and more complete when the channel answers at once.

What should you do before you build?

  1. Count. Take last week's WhatsApp orders. How many, how many lines, how many needed a call back.
  2. Check your version and licence. SQL Account version and API module, or AutoCount desktop versus cloud. One message to your dealer settles it.
  3. Clean the item list. One code per item, units stated, and a column for the names customers really use.
  4. Pick ten customers who order often and in a regular pattern. Start there.
  5. Decide who approves drafts and by what time of day.

A build like this sits next to two others we describe: PO and invoice matching on the purchasing side and e-invoice automation once the order is invoiced. The wider picture for the sector is in AI for trading and distribution in Malaysia, and the tooling in n8n WhatsApp automation.

How we help

A winsym.ai Quick Scan audit (from MYR 3,500, credited to a build started within 60 days) checks your version, your item list and one week of real orders, and tells you which import route fits. Custom builds start from MYR 12,000 and are scoped from the audit. If you first want to know whether your setup can do this at all, book a free 30-minute call.

Sources

(checked November 2026)

Frequently asked questions

Can SQL Account or AutoCount read WhatsApp messages directly?

No. Neither vendor's documentation describes a WhatsApp inbox. You need a layer in between that receives the message, turns it into order lines and writes a draft through a documented import route. That layer can be a small custom service or an automation tool such as n8n.

Does SQL Account have an API?

Yes. eStream documents a SQL Account API that uses an access key and secret key with AWS Signature v4. The on-premise setup needs SQL Account version 5.2025.1038.875 or above, a public IP address and port forwarding. The user that holds the API key can no longer log in to SQL Account directly, so create a separate API user.

Can the AutoCount Cloud Accounting API create a sales order?

The public API documentation, checked in November 2026, lists quotations, invoices, credit notes and purchase orders, but no sales order. With AutoCount Cloud Accounting we would write the WhatsApp order as a draft quotation. The desktop AutoCount Accounting has a developer wiki with sample code that creates a sales order.

Do we need the WhatsApp Business API, or is the normal app enough?

Automatic intake needs the WhatsApp Business Platform, because that is where Meta sends each incoming message to your system as a webhook. If you stay on the normal app, a semi-automatic route still works: staff forward or paste the order text into a small tool that builds the draft.

What does it cost to build?

On our price list a Quick Scan starts from MYR 3,500 and is credited to a build started within 60 days; custom builds start from MYR 12,000 and are scoped from the audit. Add your BSP or Meta messaging fees, and any charge your SQL Account or AutoCount dealer makes for the API or import module.