The AI costs about one cent. The message costs six to eight times more.

The intelligence is cheap and the messaging is not. On the accounts we operate, the AI itself meters at a bit over one US cent per action, while a single marketing message on WhatsApp costs RM 0.3467, which is about eight US cents. That is six to eight times more than the thinking that produced it.

That inversion is the whole answer, and almost nobody quotes it that way round.

One caution before the numbers, because it is the trap this page nearly fell into itself. The AI is billed in US dollars and the messages are billed in ringgit, so any ratio between them depends on an exchange rate. Every comparison below uses RM 4.04 to the US dollar, the rate on 2026-09-02. Set one cent against thirty-five sen as though a cent and a sen were the same coin and you get thirty times, which is wrong by a factor of four.

Below are the numbers, read off a live billing panel rather than off a pricing page. We have also left in the estimate we got badly wrong, because the size of the error is the useful part. And at the end, the number that matters more than any of them, which is what it costs you to not have this at all.

01 / Unit costsThe unit costs, measured

These are read from a live account's usage panel across three consecutive months in 2026.

Read these as meter readings, not as a price list. This is what the underlying usage meters at, before anyone's margin, ours included. Nobody sells it to you at cost, and an agency quoting you its own wholesale rate is either subsidising you this month or has not read its own invoice. What you would actually pay is further down, under its own heading.

The second trap is the word "action". A billed AI action is not one customer message. One exchange with one human can trigger several of them, so a bot handling a thousand conversations a month is not billing a thousand actions. Size from the panel, never from a conversation count.

02 / Our own errorWe estimated this and we were wrong by ten times

Before we had a live account big enough to measure, we sized a bot from token arithmetic: a few thousand tokens per exchange, so a fraction of a cent per message. That estimate was three to eleven times too low.

The reason is structural and it applies to every production bot, not just ours. A real bot carries a large system prompt, a knowledge base and the conversation history on every single call. Input tokens dominate. The lean FAQ exchange the estimate assumed does not exist in production.

So: never size a bot from token arithmetic. Read the billing panel.

We also have to flag something against our own numbers. The platform announced a cut of "up to 50%" to AI action costs on 2026-08-14, in the middle of the window these figures come from. Whether that reduction reached the specific meter we measured is not verified. Treat the AI column as an upper bound.

03 / Meta's lineWhat Meta charges, which is the bigger line

Malaysia moved to MYR billing on 2026-04-01. These rates took effect 2026-07-01.

Marketing has no volume discount. At any scale. This is the item most often modelled wrongly in a proposal. We checked it twice, once against the published rate card and once against the live pricing API behind Meta's own interactive calculator, which returns an empty tier list for Marketing and a full tier ladder for the other two. Utility tiers do exist, and they start above 150,000 messages a month, which is not a number any Malaysian SME will reach.

A quoted Marketing price is the final price. A thousand-recipient broadcast to Malaysian numbers is about RM 347 in Meta fees alone. Twenty thousand is about RM 6,934, and that is a floor.

These are Meta's own published rates and they are the same for everyone. They are also the only figures on this page that arrive without anybody's markup on them.

04 / A real monthWhat a real month looks like

One live account we operate, three consecutive months, everything included:

• Month one: $72
• Month two: $134
• Month three: on pace for roughly double month two

Read the shape, not the absolute. Two things drive it. First, AI spend grew faster than message spend, because more conversations means more actions per conversation, not just more conversations. Second, and more usefully:

Messaging cost scaled sub-linearly. Volume went up about 4.4 times while the messaging bill went up about 1.4 times, so the blended cost per message fell to roughly a third. The likely mechanism is that as inbound traffic grows, more of it lands inside the free 24-hour service window, where a reply costs nothing. We measured the ratios; the mechanism is our inference, not a measurement.

That is the opposite of the intuition most people bring, which is that a busy bot gets expensive fast. A busy INBOUND bot gets cheaper per message. A busy OUTBOUND one does not.

05 / The bundle trapWhat you would actually pay

Everything above is measurement. This is price, and we would rather you read it here than discover it in a proposal.

Two honest notes on that table, because they are the questions we would ask.

The unlimited AI tier is not always the right answer. It becomes the cheaper option only above a certain volume, and below that you are better off metered. On one account we run, the crossover arrived in a specific month and we moved it then, not before. Which side of the line you are on is a question with a numeric answer, and the answer changes month to month.

And the middle bundle is the one to watch. Most platforms in this category offer a mid tier that includes a fixed allowance of AI responses. On the account above, that allowance would have covered about 8% of the month's actual usage. The other 92% would have metered per use anyway, on top of the tier fee. So the ladder is not a ladder. It is metered, then a step that can cost you money, then a step that saves you money.

That is the whole reason we quote unlimited or metered and nothing in between.

06 / The real ceilingThe cost of the answer that never came

Here is the number that should decide this, and it is not on any invoice.

Uldis recently tried to book a table at a restaurant in the Klang Valley. The first reply was that someone would get back to him after 2pm. He waited about four hours. When the conversation finally started, every question he asked meant another wait, because a human on the other end was fitting him in between other jobs. The dinner, for two, came to just under RM 1,000.

He went because he had decided on that specific restaurant. Almost nobody does that. Anyone with a second choice books the second choice inside those four hours.

Then the part that makes it worse. On the night, the restaurant was somewhere between 10% and 20% full.

So the arithmetic. One table, one evening, just under RM 1,000. Unlimited AI is USD 150 a month, which at the RM 4.04 rate above is about RM 605. That single booking was worth more than the entire month's fee for the thing that would have answered him at 10am. The WhatsApp channel on top is another USD 50, about RM 202, and the two together are still less than that one dinner.

A restaurant with 80% of its tables empty cannot afford to lose the customers who are actively trying to give it money. It was not losing them to a competitor's food or price. It was losing them to a four-hour wait for a reply, which is a solved problem and has been for a while.

Read that as one story, because that is what it is. It is a single evening, a single restaurant, one person's bill, and we are not going to dress it up as a statistic. We will not tell you that businesses lose some percentage of enquiries to slow replies, because the numbers in circulation for that do not survive being traced back, and we have written about that elsewhere. What the story is good for is the ratio, and the ratio is not close: the monthly cost of answering everybody, immediately, at midnight, in the language they wrote to you in, is smaller than one average transaction in most businesses that have an average transaction worth talking about.

So count your own. Take your average transaction value. Take the number of enquiries last month that waited more than an hour for a first reply, which your inbox will tell you and which is usually a worse number than anyone expects. You do not need a conversion-rate assumption or a vendor's benchmark to see whether this pays for itself. If one recovered booking a month covers the fee, the question was never really about the fee.

07 / What to doThe thing that actually stops a big send, and it is not cost

If you are sizing a broadcast, the binding constraint is usually not the bill. It is delivery.

WhatsApp messaging limits are set per business portfolio, count unique recipients in a rolling 24 hours, and ladder in fixed steps: 250, then 2,000, then 10,000, then 100,000. Pacing is applied on top, and on a poor quality signal the remainder of a bulk send can simply be dropped without an error.

A budget that assumes 20,000 messages will go out in one day is not a budget problem waiting to happen. It is a delivery problem that looks like a budget.

08What we would tell you to do with this

Four things, in order.

1. Size the loss before you size the bill. Average transaction value times the enquiries that went unanswered for hours last month. If that number is not comfortably larger than a few hundred ringgit a month, do not buy this from us or from anybody. If it is, the rest of this page is detail.
2. Find your own numbers before you accept anyone's, including ours. Every measurement above is from accounts we run, in one market, in one quarter of 2026. Your ratio of inbound to outbound will move the total more than any vendor's per-message rate will.
3. Separate the two bills in your head. The AI is a rounding error at SME volume. The messaging is the cost, and the single biggest lever on it is how much of your traffic replies inside the free 24-hour window rather than needing a paid template to start a conversation.
4. Stop pricing outbound as though it gets cheaper in bulk. For marketing messages in Malaysia it does not, and a proposal that models a bulk discount is modelling something that does not exist.

The uncomfortable version: if the plan only works at 20,000 marketing messages a month, the plan costs RM 6,934 a month in Meta fees before it has produced a single reply, and nobody selling you the automation is going to lead with that.

The more uncomfortable version, for the buyer rather than the seller: the restaurant above is still doing it the old way tonight, and the four-hour wait costs it more per empty table than the fix costs per month.