Skip to content
AptoAir
AI usage & wallet

What AI actually costs you, per reply

AI replies are metered on the amount of text going in and coming out, and paid from a wallet you top up in advance. Typical replies cost a fraction of a rupee — and because it is prepaid, there is no such thing as a surprise AI bill.
Prepaid — never an overdraftItemised per day and modelChoose cheaper models per flow

Wallet top-up calculator

See the exact charge before you pay.

Credited to your wallet
₹1,000.00
Payment gateway fee (3%)
₹30.00
GST (18% on the fee)
₹5.40
You pay
₹1,035.40

The gateway fee and its GST are charged by the payment processor, not by us — we pass them through at cost. The full amount you asked for lands in your wallet.

How it works

Metered on text, not on time

You are not buying a seat or a monthly AI allowance. You pay for the text the model reads and writes, which means a quiet month costs almost nothing.

Input and output are measured

Every AI reply has two parts: what it reads (your brief, the conversation so far, any documents it consults) and what it writes back. Both are counted, and output is priced higher than input — that is how language models work everywhere.

Priced per model

A fast, small model costs a small fraction of what the sharpest one does. Because you choose the model per automation, a simple “what are your hours?” flow never costs what a complex sales conversation does.

Drawn from your wallet

Each reply debits your prepaid balance immediately, and appears in your usage history the same day with its token counts and rupee cost. Nothing is estimated or trued-up later.

What a real reply looks like in your usage log
FeatureCallsInOutCost
Automation AI211925₹0.0065

That is two AI calls, reading 119 tokens and writing 25, costing well under one paisa. Even at a thousand conversations a month, a simple FAQ agent on the fastest model costs less than a cup of coffee. Longer briefs, deeper history and sharper models cost more — which is why the choice is yours, per automation.

Models

Pick the cheapest model that does the job

Three tiers, chosen per automation. Most teams use the fast one for routing and the balanced one for real conversations.

Fastest

Cheapest

Good for

Routing, greetings, simple FAQs, “are you open?”

Trade-off

Very cheap and very fast. Struggles with long, multi-step reasoning.

Balanced

Default

Good for

Standard customer conversations — the sensible starting point.

Trade-off

Reliable for most support and sales chat, at a modest cost.

Sharpest

Most capable

Good for

Long conversations, careful instruction-following, nuanced sales.

Trade-off

Noticeably better at staying on brief. Costs more per reply.

Model names inside the product are ours, not the underlying vendor's. The point is the trade-off, not the badge — and you can change it on any automation at any time.

Control

You cannot overspend by accident

The wallet is prepaid on purpose. There is no credit line, no overage rate and no invoice arriving with a number you did not expect.

  • Top up any amount, whenever you want
  • A low-balance email before you run out
  • Optional auto top-up if you would rather not watch it
  • When the balance hits zero, AI replies pause — everything else keeps working
  • Every debit itemised by day, feature and model
  • Gateway fee and its GST passed through at cost, shown before you pay

Five ways to spend less

  1. 1Keep your system brief tight. It is read on every single reply, so a rambling brief costs you on every message.
  2. 2Reduce history depth. Reading the last 5 messages instead of 30 cuts input dramatically with little loss for most support chat.
  3. 3Cap reply length. Customers prefer short WhatsApp answers anyway.
  4. 4Use the fastest model for routing and greetings, and reserve the sharper one for conversations that actually need it.
  5. 5Handle the obvious cases with plain automation blocks. A keyword that sends your price list costs nothing at all.
Questions

Frequently asked

No. If your automations contain no AI block, your wallet is never touched. The inbox, campaigns, templates, contacts and store integrations all work without any AI spend.
Roughly three-quarters of a word in English. “Your order has shipped” is about five tokens. Both what the model reads and what it writes are counted, which is why a long brief and deep history cost more per reply than a short one.
No, they are separate. The subscription is the software fee, billed monthly or yearly. The wallet is prepaid credit used only for AI usage. Having one does not top up the other.
AI replies stop. Your inbox, campaigns, automations without AI blocks, order updates and everything else carry on exactly as normal. You get an email warning before this happens.
The payment gateway charges a percentage to process the card or UPI transaction, and GST applies to that fee. We pass both through at cost and show the exact breakdown before you pay — the full amount you asked for is what lands in your wallet.
Wallet balance is prepaid credit for usage. See the refund policy for how this is handled, or ask us before topping up a large amount.

Try the AI on your own questions

Start the trial, paste in your FAQ, and message your own number. You will know within ten minutes whether it is good enough for your customers.

No credit cardCancel anytimeYour data stays yours