WhatsApp API replies stop being free on 1 October

Replying to a customer stops being free on 1 October for any South African business running WhatsApp through the Business Platform, and Meta has not yet published what it will cost. The number that decides your bill is not Meta's price. It is how many messages your business sends to close one job.

TechCentral reported on 19 August that Meta would not confirm the price, and set out the mechanic. An hour-long exchange that today produces one billable message will, from October, produce five charges. Meta's own pricing documentation confirms the change takes effect on 1 October and commits to announcing per-market rates no later than 1 September.

So there is no price yet, and building a rand model would be dishonest. The estimates circulating locally are reverse-engineered rather than published. There is a mechanic, though, and it is per message, which makes the exposure countable today.

The scope matters as much as the price. This is the WhatsApp Business Platform, the API that businesses reach through a provider. A plumbing firm in Boksburg running the free WhatsApp Business app on a handset is not billed for replies. The moment that firm moves onto a platform to get a shared inbox, templates or automation, it is.

The number Meta has not published, and the one you already own

Whatever the charge turns out to be, it multiplies against a quantity that belongs entirely to you: how many messages it takes to resolve one job.

That quantity is measurable today, for free, on a phone you are already holding. Nobody needs Meta's price to do it. Open your last twenty customer threads and count the messages from the first contact to the point where the customer stopped replying.

Then count how many of those messages were your staff asking the customer for something the business already had on record.

That second number is the one that will start costing money. It is also the only one you can change before October.

Where the extra messages come from

A business whose job cards, customer history and invoicing sit in three separate systems re-asks the customer for the address, the reference and the balance. The technician in the WhatsApp thread cannot see the accounting package. The person who can see it is not in the thread.

So the conversation goes out and comes back three times to recover data the business already owns. Under the current arrangement that costs nothing but goodwill. After October it appears on an invoice.

The common shapes are worth naming, because they are all fixable and none of them are a WhatsApp problem:

• The delivery address is in the CRM, but the driver works from a chat thread
• The invoice reference is in the accounting package, which field staff cannot open
• The customer's balance sits on a statement nobody in the conversation can see
• The last job's notes are in a different technician's phone

Each of those is one avoidable round trip. Four round trips on every job, across two hundred jobs a month, is eight hundred paid messages you are choosing to send.

What deflection looks like when someone has paid for it

Absa has said that its chatbot fields more than 100 000 queries a month, alongside more than 1 400 developers using AI coding tools. That is a bank with a budget for the problem, and it is a useful upper bound rather than a model to copy.

A forty-person distribution business is not building that. What it can do is smaller and closer to hand. Most repeat questions in a small business are the same four questions, and the answers to all four already exist in systems the business pays for every month.

What blocks this in South Africa

Three things, and none of them are technical.

WhatsApp is not optional here. Telling customers to email instead is not a cost saving, it is a churn plan. The channel stays. Only the message count is negotiable.

POPIA sits on the obvious fix. Pulling a customer's history into a chat thread so staff stop asking for it means personal information moving between systems and onto personal handsets. That needs a lawful basis, a retention position and a record of who can see what. It is workable. It is not something to discover after the fact.

Field staff use WhatsApp because it survives. When the office connection is down, the phone in someone's pocket still works. Any answer that depends on everyone sitting in front of a desktop system will quietly be ignored and the thread will carry on regardless.

The fix is joining up, not replacing

The instinct will be to buy a customer service platform, and for most businesses this size that is the wrong spend. The information is not missing. It is scattered across an accounting package, a spreadsheet of jobs and four phones.

Connecting those so the person in the conversation can see the record is a smaller piece of work than replacing any one of them. It also does not ask staff to change the channel they already use. This is the kind of automation that pays for itself in message count rather than in headcount, which is why Zorah starts by auditing where the answers already live.

The businesses that will feel this change in October are the ones sending five messages to do one thing. The businesses that will not are the ones sending two.

What to do on Monday

Open WhatsApp and pick your twenty most recent customer threads. For each one, write down two numbers: total messages, and messages where your side asked for information the business already stored somewhere.

Add up the second column. Multiply by your monthly job volume. That is the exposure, expressed in the only unit that matters, and you have it six weeks before anyone pays for it and two weeks before Meta publishes a rate to multiply it by.

Then check one more thing. Find out which staff phone numbers your customers actually message. If half your customer conversations are happening on a technician's personal number, the cost is the smaller of your two problems.