AI usage and spend
What AI agents cost, what you're actually charged for, and the ways to make sure the bill can't surprise you.
AI agents are the one genuinely variable cost in Enscale. Everything else is your plan; this is charged per use.
The reassuring part: the rate is flat and tiny, and there's a spending limit you control.
What you pay
$0.02 per AI reply — voice notes and images included, the same on every paid plan.
Not available on the Free plan, and Free is never billed.
What counts as an AI reply
Every reply your AI agent sends to a customer is charged, whether the conversation ends resolved or is handed to your team.
| What happened | Charged? |
|---|---|
| The AI sent a reply | Yes — $0.02 |
| The AI replied to a voice note or image | Yes — $0.02, nothing extra |
| The AI's holding message, fallback line, or "please send text" reply | No |
| A follow-up nudge the AI sends | Yes, $0.02 |
| A reply your team sent | No |
| The AI handed the conversation over | Nothing extra — earlier AI replies are charged |
Handing over costs nothing extra
Handing a conversation to your team is not a separate charge. The AI replies already sent in that conversation are charged like any other, but handing over itself is free.
Tune your agents towards handing over readily. An agent guessing at your refund policy to avoid escalating is far more expensive than one extra handover.
Voice notes and images are included
If you turn on audio or image understanding, the agent can answer voice notes and product photos — normal on WhatsApp in India and much of the Middle East. It costs nothing extra: a reply is $0.02 whether the customer typed, spoke or sent a photo.
Free AI during a plan's trial days
During a plan's Shopify trial days, the first 100 AI replies are free. After that, AI conversations go to your team until the trial ends, and usage is billed normally afterwards.
Three ways to control spend
1. Set a monthly spending limit — your hard ceiling
In Settings → Billing you can set one optional monthly limit on your total usage charges — AI replies and any overage combined.
Nothing is billed past it. Once the limit is reached, AI conversations are handed to your team and overage turns off. It lifts when you raise or remove the limit, or when the calendar month rolls over. Usage held back by the limit is never billed in a later month.
This is the setting that makes AI spend genuinely safe. It cannot run away from you.
2. Point agents at the right conversations
The cheapest reply is one that never needed to happen.
- Enable agents where they earn their keep. Webchat pre-sale questions are high-volume and highly automatable. A quiet Facebook Page probably isn't worth it.
- Use automations for anything predictable. An order-status automation costs nothing; an AI agent answering "where's my order?" costs replies and might get it wrong.
- Keep your knowledge base tight. An agent that flails through six turns before answering costs more than one that answers in two.
3. Watch the spend
Settings → Billing shows AI replies this period, your projected charge, and where you sit against your limit.
Worked example: what this actually costs
You're on Growth ($0.02 per AI reply, the same as every paid plan) and your webchat agent handles about 400 conversations a month, averaging about 6 AI replies each.
| Conversations the agent handled | 400 |
| Of those, handed over to your team | 120 — their AI replies are still charged |
| Of those, involving a voice note | 60 — no extra charge |
| AI replies sent | 400 × ~6 = 2,400 |
| Monthly AI cost | 2,400 × $0.02 = $48.00 |
If you'd set a $100 monthly spending limit, you're using under half of it.
Is it worth it? About 12 cents a conversation is cheaper than almost any human alternative.
But the real question isn't cost per reply — it's whether the replies were any good. One badly-handled conversation that becomes a public complaint costs far more than the cents you saved. Which is the whole argument for letting the agent escalate freely.
Keeping an eye on it
The number worth watching is replies per conversation — total AI replies divided by conversations handled. If it's creeping up, your agent is taking more turns than it should, usually a knowledge base gap.
Track the value side on the Revenue page under support savings — and read it next to your AI charges, never instead of them.
Common questions
Next steps
- AI agents — configuring what agents handle.
- Knowledge base — fewer turns per conversation.
- Billing and usage — how this appears on your Shopify invoice.