The rules that decide the design
This is not a chat widget with a green logo.
WhatsApp is not an open channel you can build whatever you like on. Three platform rules shape every decision, and a build that ignores them either gets rejected or quietly stops delivering — usually after the invoice.
You get 24 hours, then you get templates
When a customer messages you, you can reply freely for 24 hours. After that you may only send message templates that Meta has approved in advance, and you cannot improvise. Almost every disappointment with this channel traces back to that one rule: the abandoned-cart nudge, the appointment reminder, the delivery update and the win-back all live outside the window and all have to be written, submitted and approved before they can ever be sent. Design for it up front and the channel is excellent; discover it in week six and half the plan is gone.
Templates are reviewed, categorised and rated
Every template goes to Meta for approval and lands in a category — marketing, utility or authentication — and the category is what you are billed against, so a badly categorised template can cost several times what the same message should. On top of that your number carries a quality rating: if people block or report you, the rating falls and your sending limits fall with it. Marketing volume is therefore something to earn, not something to switch on, and anyone promising blast campaigns on this channel is describing how to get a number rate-limited.
Opt-in is the platform's rule, not a legal footnote
You must have permission before you message someone, and you must be able to show where it came from. That is a real constraint on a list bought or scraped from anywhere, and it is a reason a good build starts by adding opt-in to the places people already reach you — the checkout, the contact form, the QR code on the counter — before it sends anything. It also means the useful metric in month one is opted-in contacts, not messages sent.
What a build actually contains
Six pieces, and the bot is only one of them.
Channel and audience check
Before anything is built: whether the people you sell to use this app at all, and what they use instead if they do not. This is free and it is the step most likely to end the conversation.
Number, verification and ownership
Business verification, a number registered to your company, and the assets in your own Meta Business Manager rather than ours.
Conversation design and handoff
The flows, the fallbacks, and the point where a person takes over. A bot that cannot hand off is a complaint generator.
Templates written and submitted
The out-of-window messages drafted, categorised correctly, submitted for approval, and revised when review pushes back.
Integration with what you already run
Orders, bookings, stock and CRM records read from the system that owns them, so the bot answers from the truth instead of a stale copy.
Logs, review and tuning
Transcripts you can read, a weekly look at what the bot got wrong, and the rules corrected. Automation that nobody audits drifts.
Who this is genuinely right for
In Pakistan the question barely arises — it is how businesses and customers already talk, and a company answering enquiries by hand on a personal number is the norm rather than the exception. In Britain it is close behind and it is the default for anything involving a diaspora or an international supply chain. In the United States it is a different picture: most consumers are on SMS and iMessage, and building here for a domestic American consumer audience is usually the wrong call.
That does not make it wrong for every American business. The ones it works for are specific and easy to recognise: companies selling into Latin America, India, the Middle East, Africa or Southern Europe, where it is the dominant channel; businesses serving immigrant and diaspora communities at home; anyone with suppliers, drivers, contractors or field staff abroad who already coordinate on it; and travel, logistics and education businesses whose customers are outside the country while they need support most. If you are one of those, this is a serious channel. If you are a US brand selling to US consumers, we will tell you to spend the money on SMS or on your existing support desk, and that advice costs you nothing.
Where AI belongs in this, and where it does not
Most of what pays back on this channel is not a model. It is a rule that fires at the right moment: an order confirmation, a reminder the day before, a delivery update, a follow-up when a form is abandoned. Rules are cheap, they are testable, and they do not invent an answer at midnight. The place a model genuinely earns its keep is the messy middle — understanding a question asked in three languages and a spelling nobody standardised, summarising a long thread for the human taking over, drafting a reply an agent then approves. We build both, we price them differently, and we tell you which one your problem is before you commit to either.
What we are not going to show you
Plainly: there is no published case study behind this page. The automation work sits inside client systems and has not been written up, and our own internal build is not something we would ask you to take on trust. So there is no percentage on this page, because we do not have one you could verify. What there is instead is a free read of your actual process — where enquiries arrive, what fraction are the same four questions, and what a rule would fix before anything is bought. If another agency shows you a number, ask which client, over what period, and whether you may phone them.
How it runs
Fit first, then a small live scope, then more.
Fit check, free
Who your customers are, whether they are on this channel, and what the highest-payback conversation is. Written, and sometimes it ends here.
Number, verification and access
Business verification started early because it is the slowest step, the number registered to you, and the platform assets created in your own account.
One flow live, end to end
The single highest-volume conversation built, integrated, templated and launched — not eleven flows half-finished.
Read the logs, then widen
Real transcripts reviewed weekly, the rules corrected, and the next flow added only once the first one is behaving.
Money
Two costs, and one of them is not ours.
Fit check
The channel and audience read, written down. Free, delivered whether or not it concludes in our favour, and yours to take elsewhere.
- ✓No commitment
- ✓Written, not a call
- ✓Says no when the answer is no
First flow, fixed scope
One conversation designed, built, integrated, templated and launched, with the account set up in your name. Quoted once from the flow, not by the hour.
- ✓Verification handled
- ✓Templates written and submitted
- ✓Your Business Manager
Ongoing
Monthly, for the transcript reviews, template changes when review rejects one, and new flows as they earn their place. Optional and cancellable.
- ✓Weekly log review
- ✓Template maintenance
- ✓Cancel with notice
No figures appear on this site, in any currency. Note also that Meta bills you directly for messages sent outside the 24-hour window, priced by template category and destination country — that cost is yours, it is not ours to mark up, and any proposal that hides it inside a single monthly number is worth questioning.
FAQ
What people ask before starting.
More from GOFTECH
Where to look next.
Flutter App Development
A Flutter app development company for the USA, the UK and Pakistan, working from one office in Islamabad. Six Flutter apps live on both…
Build — Products that ship
Build with GOFTECH: mobile apps (Flutter), custom web systems (Laravel & React), Shopify and WordPress stores, AI automation and UI/UX…
AI & Workflow Automation in Pakistan
An AI automation agency in Pakistan building WhatsApp order confirmation, lead routing and document assistants — with an honest line…