How much AI does a business like yours use?
In short: far less than it seems, because the button menu handles almost everything you get asked, and that costs nothing. A bot that sends every message to the AI can cost up to 35 times more per conversation than one that saves the AI for what isn't on the menu. Below are three example businesses with the numbers, and further down the full calculation if you want to check it.
Three businesses, and how much AI each one uses
The math is always the same: out of everything people write to you, whatever is on your menu gets answered without spending AI, and only the rest reaches the AI. The better your menu, the less you spend.
Restaurant
300 orders a month
The menu, opening hours, delivery area and order taking all fit in the button menu. Only one-off questions reach the AI: whether a dish has gluten, whether a side can be swapped.
Online store
1,200 inquiries a month
Order status, shipping and returns are menu. What goes off script are questions about a specific product and comparisons between two models.
Clinic
500 appointments a month
Booking, reminding and rescheduling is menu, and so are insurers. Anything clinical deliberately isn't handled by the AI: it goes to a person.
With those numbers you can compare against the AI messages each plan includes in the pricing table. One thing to keep in mind: a conversation handled by AI isn't one message, it's several — so if your business looks like the one on the right, also look at the option to connect your own API key, which has no cap.
What do you really pay for 1,000 conversations?
In short: a bot that sends everything to the AI costs up to 35 times more per conversation than one that lets the button menu handle the usual questions and saves the AI for questions that aren't on the menu.
| Conversations/month | Pure GPT-4o | Pure GPT-4.1 | FlujosChat hybrid | JSON flows only |
|---|---|---|---|---|
| 1,000 | $3.50 | $4.20 | $0.78 | $0.10 |
| 10,000 | $35 | $42 | $7.80 | $1.00 |
| 50,000 | $175 | $210 | $39 | $5.00 |
| 100,000 | $350 | $420 | $78 | $10 |
See the calculation's assumptions
- Average conversation: 2 turns, ~600 input tokens + ~200 output tokens
- Hybrid mode: 80% handled by the button menu, 20% sent to the AI
- GPT-4o: $2.50 per million input tokens, $10.00 output (April 2026 prices)
- GPT-4.1: $3.00 per million input tokens, $12.00 output
- Flows only: the only cost is the server, ~$10/month, with no per-message cost
Why most of your conversations don't need AI
Look at the history of your own WhatsApp: almost everything is a variation of the same twenty or thirty questions. There, the AI adds nothing the menu can't do just as well — it only costs money and can get wrong a piece of data that had to be exact.
The menu handles this, without spending AI
- ✓"What are your hours?"
- ✓"How much is product X?"
- ✓"How do I place an order?"
- ✓"I want to cancel my subscription"
- ✓"Do you ship to [city]?"
- ✓"Do you have any promotions?"
- ✓"Phone number / address"
- ✓"I need an invoice for my purchase"
And this is worth leaving to the AI
- AIA complaint with several things mixed together
- AIComparing two products the customer names
- AIMessages that ask two things at once
- AIWhen the customer writes the way they talk, with slang
- AI“So which one do you recommend for my case?”
- AIWhen they're upset and you need to understand why
Where these numbers come from
None of this is a rough guess on our part: it's OpenAI's public rates multiplied by the volume. Here's the full calculation, in case you want to check it or redo it with your own numbers.
See the detailed calculation↓
A typical repetitive conversation takes about 600 “pieces of text” of input (what the AI reads: your question plus the context) and about 200 of output (what it replies). OpenAI charges for those pieces, and output costs four times more than input.
At 50,000 conversations a month, sending them all to the AI, input comes to about $75 and output to about $100: $175 a month in AI alone. If the menu handles 80%, you only pay for the remaining 20%: $39 a month, 78% less.
- Average conversation: 2 turns, ~600 input tokens + ~200 output tokens
- Hybrid mode: 80% handled by the button menu, 20% sent to the AI
- GPT-4o: $2.50 per million input tokens, $10.00 output (April 2026 prices)
- GPT-4.1: $3.00 per million input tokens, $12.00 output
- Flows only: the only cost is the server, ~$10/month, with no per-message cost
And to be upfront: this is what the AI would cost if you built the bot yourself. On a FlujosChat plan the AI is included, and what you pay for isn't the AI — it's the whole platform: getting your number approved with Meta, the dashboard, the flows, your business's team and payments.
Start saving today
7 days free. No card. Set up in 5 minutes. The hybrid plan is included from the Sales AI tier.