E-commerce & Retail

WhatsApp automation: reminders, receipts and replies that run themselves

Official WhatsApp Business API automation — payment reminders, order updates and campaigns, with an AI assistant on the replies.

Industry: E-commerce & Retail
Platform: AI

In India, WhatsApp is where customers actually are — and where most businesses still work manually, one message at a time, until the volume makes it impossible. This is our reference build for doing it properly on the official WhatsApp Business API: approved message templates for reminders, receipts and order updates; campaign sends with consent and opt-out handled correctly; and an AI assistant that answers the replies, because an automated message that provokes a question nobody answers is worse than no message.

The challenge

WhatsApp automation is deceptively strict. The official API only permits template messages outside a 24-hour customer service window, templates need approval, and opt-out has to be honoured or the number gets restricted. Getting this wrong does not degrade gracefully — it ends with the business's number blocked. The second problem is the replies. Automated messages generate inbound conversation at exactly the volume that made manual work impossible in the first place.

Our strategy

Treat the platform's rules as the design constraints they are: templates modelled as first-class objects with approval state, sending windows enforced in code rather than by convention, and consent recorded per contact with opt-out honoured everywhere automatically. Then close the loop by putting a grounded AI assistant on the inbound side, so the replies an automated reminder generates get answered too.

The solution

Scheduled and event-driven messages — payment reminders, order and delivery updates, appointment confirmations, renewal notices — sent through approved templates on the official API, with delivery and read status recorded against each contact. Campaigns segment the customer list, respect opt-out, and never send outside permitted windows. Inbound replies reach an AI assistant grounded in the business's own information, with clean handover to a human for anything it should not handle.

Results

What changed

Official

Cloud API

The compliant route

Built on Meta's official API, so the business number stays in good standing.
0

manual sends

Reminders that run themselves

Events in your system trigger the message; nobody has to remember.
2-way

conversation

Replies answered too

An AI assistant handles the inbound side, escalating when it should.

Key features

What we built

Official Business API

Built on the WhatsApp Cloud API — not an unofficial bridge that gets numbers banned.
Template lifecycle managed

Approval state modelled in the system, so unapproved templates cannot be sent.
Event-driven reminders

Payments, renewals, appointments and orders trigger their own messages.
Campaigns with consent

Segmented sends with opt-out honoured automatically everywhere.
AI on the replies

A grounded assistant answers inbound questions, with human handover.
Delivery and read tracking

Per-contact status from verified webhooks, applied idempotently.

Frequently asked questions

Is this a client case study?

No — it is a capability showcase describing our WhatsApp automation reference build. We deploy and adapt it per client; there is no client name or live URL attached to the showcase itself.

Do you use the official WhatsApp API?

Yes, exclusively — the official Cloud API. Unofficial bridges are cheaper right up until the business number is restricted, at which point the channel is gone along with the customer list's trust.

How is opt-out handled?

Consent and opt-out are stored per contact and enforced at the sending layer, so an opted-out contact is unsendable by construction rather than by someone remembering to filter a list.

Can it plug into our existing system?

That is the usual case. The automation reads events from your existing platform over an API — orders, invoices, appointments — and sends against them. Our own products, like GoLibora and SchoolDarpan, use exactly this pattern internally.