Connecting WhatsApp to Business Software: What to Plan
To automate WhatsApp for a small business, decide what each message is for before anyone writes code. Separate messages your software sends automatically from conversations a person handles. Record how each customer agreed to hear from you, plan for messages that need approval before use, and name who watches for failed sends. Link every message to the order or record it concerns.
The rest of this guide turns those decisions into questions you can work through with your team or build partner.
Which messages are system-triggered and which are conversational?
Start by listing every message you want to send, then sort each one into one of two groups.
System-triggered messages come from an event in your software: an order is confirmed, a delivery is scheduled, a payment is received. Nobody types them. They follow a fixed structure with a few fields filled in, such as a name or order number.
Conversational messages come from a person. A customer asks about stock, a staff member answers. These need someone to own the inbox, a response expectation and a way to hand over between team members.
Most problems start when the two groups are mixed up. If an automated confirmation invites a reply and nobody is watching for replies, customers feel ignored. Decide for every automated message what happens when the customer replies, even if the answer is "a person picks it up during business hours".
Use a flow table like this one to make the decisions visible:
| Business event | Message type | Who or what sends it | If the customer replies | Linked record |
|---|---|---|---|---|
| Order confirmed | System-triggered | Order system | Routed to sales inbox | Order |
| Dispatch scheduled | System-triggered | Dispatch process | Routed to delivery team | Order and shipment |
| Payment received | System-triggered | Billing process | Routed to accounts | Invoice |
| Product question | Conversational | Staff member | Same staff member continues | Customer, optional order |
| Complaint | Conversational | Support lead | Escalation route applies | Order and support case |
The rows above are illustrative. Your list will reflect your own processes.
How should consent and opt-in work?
Treat consent as a principle built into the system, not a box ticked once. Only message people who have agreed to hear from you on this channel, for the kinds of messages you intend to send.
In planning terms, that means your software should store:
- Whether the customer has opted in, and for which message types
- Where and when the agreement was given, such as a checkout form or a counter request
- How the customer can stop messages, and that the stop request is honoured everywhere
Keep operational messages, such as order updates, distinct from promotional ones. A customer who wants delivery updates has not necessarily agreed to offers. Platform rules and local law set the specific requirements; confirm those with the provider's current documentation and your own adviser rather than assuming.
Consent data is personal data. Decide who can see it and how long it is kept, alongside your wider data retention and deletion policy.
Why should template approval be treated as a process?
Business messaging platforms generally require pre-defined message formats for messages the business starts, and these may need review before they can be used. Rather than learning the exact rules mid-build, plan the workflow around them.
Treat each template like a small piece of content with an owner:
- Draft the wording and the fields it will fill in.
- Agree internally who signs off the text.
- Submit it for review and record the outcome.
- Store the approved version identifier in your software, not just the text.
- Have a fallback if a template is rejected or later withdrawn.
Changing a word in a live template may mean another review cycle. Build that into release planning so a wording change does not block a launch. Keep a simple register of templates, their status and which business event uses each one.
Who watches for failures?
Messages fail. A number may be wrong, a customer may have blocked the business, a template may be paused, or the provider may be slow. If nobody is watching, the first sign is an annoyed customer.
Plan these points before launch:
- Delivery status. Providers typically report status changes back to your system. Receiving those updates as events is usually better than repeatedly asking for them; see webhooks vs API polling for the trade-off.
- Sending limits. Expect limits on how fast you can send. Queue messages and back off when told to slow down, as covered in how to handle API rate limits.
- Retries without duplicates. A retried send should not deliver the same confirmation twice. Ask how repeats are prevented.
- A named owner. Someone sees a daily list of failed messages and decides whether to resend, correct the number or contact the customer another way.
Sending should happen in the background, so a slow provider never stops a staff member from saving an order.
How do messages link back to orders?
Every message your system sends or receives should be traceable to a business record. Without that link, staff search through chats to answer simple questions such as "did this customer get their dispatch notice?"
Store, for each message:
- The order, invoice or case it relates to
- The template and version used, if any
- Send time, delivery status and any failure reason
- Who sent it, for conversational messages
Incoming replies are harder. A customer may reply to an old confirmation about a new order. Decide whether replies attach to the most recent open order, to the customer record, or to a queue for a person to sort. Showing the message history on the order screen saves staff time and reduces mistakes.
For the wider set of questions any outside connection raises, such as authentication, testing and provider changes, read what to plan for in a third-party API integration.
Planning checklist
Work through this list before the build starts:
- [ ] Every message is listed and marked system-triggered or conversational
- [ ] Each automated message has a defined reply route
- [ ] Opt-in is stored per customer and per message type
- [ ] Opt-out is honoured across all sending paths
- [ ] Each template has an owner, sign-off step and fallback
- [ ] Sending runs in the background with queueing and retry rules
- [ ] Delivery status updates are received and stored
- [ ] A named person reviews failed messages
- [ ] Every message links to an order, invoice, case or customer
- [ ] Testing uses test numbers, never real customers
What are the common mistakes, and when does this not apply?
Automating everything at once. Start with one or two high-value messages, such as order confirmation, and expand once failure handling works.
Sending from a personal phone alongside the system. Conversations split across devices cannot be traced or handed over.
Hard-coding message text. When text lives in code, every wording change needs a developer. Keep templates as managed data.
Ignoring replies to automated messages. Customers treat any message as a conversation.
Testing on real customers. Use dedicated test numbers and a separate environment.
This level of planning may be more than you need if you send a handful of messages a day by hand from a business app with no link to other software. It becomes worthwhile once messages are triggered by your order, billing or delivery systems, or once more than one person answers customers.
Frequently asked questions
Can we start with manual messages and automate later?
Yes. Many businesses start manually to learn which messages customers value. Keep a simple record of what you send and why, because that list becomes your flow table. When you automate, migrate one message type at a time and confirm failure handling works before adding the next.
Do we need a developer to connect WhatsApp to our software?
For messages triggered by your own order or billing system, usually yes. Someone must connect events in your software to the messaging provider, handle status updates, store consent and link messages to records. Ready-made connectors can help with common tools, but check they cover failure handling and traceability.
What happens if a template is rejected?
The message linked to that template cannot be sent in that form until a revised version is accepted. This is why each template needs an owner and a fallback, such as email or a staff follow-up. Review the feedback, adjust the wording and resubmit through your normal sign-off step.
How do we stop customers getting duplicate messages?
Give each send a unique reference tied to the business event, such as an order confirmation for a specific order. Before sending or retrying, the system checks whether that reference was already delivered. This prevents repeats when a provider times out or a background job runs twice after a failure.
*Planning to connect messaging to your order or billing software? Start a conversation →*