AI usage: what each plan includes
Each plan includes a monthly AI usage allowance, and that’s what changes from one plan to the next: Personal is the base, Essential includes 5× that usage and MAX, 20×. Everything else — the attendant, the store, the history and the price of credits — is the same in all three.
What a token is
A token is a piece of text the attendant reads and writes in order to answer: the customer’s question, what it looks up in your catalog and your orders, and the reply it sends. A long conversation uses more than a short one.
The count adds up what the attendant reads and what it writes. Most of it is reading: to answer well, it reads the conversation so far and what it looked up in your store, and only then writes the reply.
The AI model
Andavor’s AI attendant currently runs on the AN-01 model. It is the one that reads the conversation, looks things up in your store and writes the reply, and its usage is what the plan allowance and the usage credits measure.
How much each plan includes
| Plan | AI usage per month | Compared with Personal | Conversations per month, about |
|---|---|---|---|
| Personal | 30 million tokens | 1× | 96 |
| Essential | 150 million tokens | 5× | 480 |
| MAX | 600 million tokens | 20× | 1,900 |
| Enterprise | Tailored | — | — |
The allowance applies to the calendar month — from the first to the last day of the month — and starts over when the month resets. On Enterprise, it’s sized to your operation and agreed with you.
Where “about” comes from
The conversations column translates tokens into conversations, and the translation comes from real customer service: in a store running on Andavor, we measured 312,460 tokens per conversation on average. A conversation here is the entire exchange with a customer — not each reply.
That’s why the page says “about” and rounds down. A long conversation, with several catalog lookups, uses more than the average; a quick question uses less. The average is there to help you choose a plan — it’s not a guarantee of how many conversations fit in your month.
When a month needs more
Once you go past the allowance, the attendant keeps answering with usage credits, at R$1.00 per million tokens read and R$8.00 per million written. If a conversation goes through without token measurement, it is charged at the per-response price of R$0.05. These prices are the same on all three plans.
- Buy whenever you want, right from the dashboard
- Pay by card or boleto (Brazilian bank slip)
- Optional automatic top-up, with a monthly cap you choose
If the month’s allowance runs out and there are no credits, automatic replies stay paused until you buy credits or until the month resets — and the customer is told right away, instead of being left talking to no one. The dashboard warns you before it gets to that point.
If your month often goes past the allowance, it’s worth looking at the next plan up: the AI usage included grows from one plan to the next.