AI Voice Minutes vs Network Minutes: What Businesses Are Actually Paying For

Here is a bill that confuses people, and reasonably so.

You buy an AI phone product with "4,000 minutes included". A month later you have a subscription charge, a separate carrier charge, an overage line you did not expect, and a total roughly double what you planned. Nobody lied to you. You were quoted one of three different things and billed for all three.

This article separates them. It is worth ten minutes because the distinction survives whichever vendor you choose.

Three things called "minutes"

1. Calls inside your own company

A member of staff dials a colleague on their extension. Or dials one of your own AI Employees to ask it something.

What happens technically: nothing leaves your organisation. No telephone carrier is involved, because there is no telephone network in the path — it is two endpoints inside the same system.

What it should cost: nothing in calling minutes. Nobody carried anything anywhere.

Why it matters more than it sounds: if asking your own AI Employee a question is metered like ringing a customer in another country, your team will use it sparingly and slightly resentfully. The whole point of giving an AI Employee an extension is that people dial it as readily as they dial a colleague. Metering that behaviour discourages exactly the thing you paid for.

When comparing vendors, ask directly: are calls between our own staff and our own AI Employees charged? Some meter everything because it is simpler for them to build. That is a real cost passed to you for their convenience.

2. Network minutes

A customer dials your number. Or your business dials a customer.

What happens technically: a telephone carrier carries that call across the public network. This is genuine third-party infrastructure and there is a genuine third-party cost, which varies by destination country and by whether the number is a landline, a mobile or toll-free.

What it costs: real money that nobody can make free, because a carrier is charging for it.

How it should be billed — and this is where models diverge.

Included allowance: "your plan includes 4,000 minutes". Feels generous. Fails quietly. You cannot see the allowance depleting, you have no natural signal as you approach it, and the month you exceed it is the month you get a bill you did not plan for. Worse, the overage rate is usually the least-shopped number in the whole quote.

Prepaid balance: you top up an amount and it draws down. Less exciting, considerably easier to control. You can see it. Running low is visible before it is expensive. And it is not entangled with your subscription, so a busy month does not silently become an expensive one.

LUIXER uses the prepaid model deliberately, which is why no plan on our pricing page advertises a minute allowance. That absence is the design, not an omission.

3. AI voice participation

An AI Employee listens, understands and speaks on a call.

What happens technically: speech is transcribed, a language model works out what to say, and a reply is spoken back — continuously, in real time, for the duration of the conversation. That is computation, and it is a different resource from the carrier leg entirely.

What it costs: measured separately from network minutes, because it scales with AI involvement rather than with call routing.

The clarifying detail: a call between two humans on your team uses no AI voice at all, even if it uses network minutes. A conversation between a customer and an AI Employee uses both. A colleague dialling your AI Employee internally uses AI voice but no network minutes.

They are genuinely independent, which is why a single "minutes" number cannot describe them.

Why one number cannot work

Consider two businesses with identical call volumes.

A professional services firm. Heavy internal collaboration, modest external volume, and an AI Employee handling reception. High internal usage (should be free), moderate network minutes, moderate AI voice.

An outbound sales operation. Almost no internal calling, very high external volume, AI Employees on most conversations. Low internal, very high network minutes, very high AI voice.

Any pricing model that lumps all three together will overcharge one of these and undercharge the other. Usually it overcharges the first, because internal collaboration is invisible in a single aggregate and gets billed at the external rate.

If a vendor quotes you one minutes figure, ask which of the three it covers. The answer is often "we hadn't separated them", which tells you what your bill will do.

What to ask before you sign

1. Are calls between our own staff, and to our own AI Employees, charged? At what rate?
2. Are network minutes included in the plan or a separate balance?
3. If included — what is the overage rate, and what happens when we hit the limit? Do calls stop, or do we get a bill?
4. If prepaid — can we see the balance? Do we get warned before it runs out?
5. Is AI participation billed separately from carrier traffic? How is it measured — per minute, per second, per conversation?
6. Are inbound and outbound the same rate? (Often not.)
7. Which destinations cost more, and can we cap or restrict outbound calling?

Question 7 is worth pressing on. Outbound calling to expensive destinations is where AI phone bills go genuinely wrong, and the protection is a default-deny policy: nothing can be dialled unless you have allowed it, rather than everything allowed until someone notices.

A note on "unlimited"

Nobody can offer unlimited calling to the public telephone network, because carriers charge per minute and no vendor absorbs that indefinitely. "Unlimited" in this market means one of three things: a fair-use cap you will discover later, a restriction to certain destinations, or unlimited internal calling described ambiguously.

The third is legitimate and genuinely valuable — internal calls really are free, because nobody carries them. But it should be stated as what it is rather than as a headline that implies the carrier traffic is free too.

The short version

• Inside your company: no calling minutes. Should be free, because no network is involved.
• To and from the outside world: network minutes. Real carrier cost. Better as a prepaid balance you can see than an allowance you cannot.
• AI taking part: measured separately again, because it scales with AI involvement rather than routing.

Three resources, three behaviours. A quote that gives you one number is not simpler — it is less informative, and the difference shows up on the bill.

See how the plans work, or how LUIXER Calling works.

Talk to us about your call volume →