If you run an answering service, a telephony company or an IT firm, your clients are already asking: can't that phone be answered by AI? Reselling is possible. The more interesting question is what lands on your plate the moment your logo is on it.
Because that is the real difference between white label and referral. With a referral it is someone else's client. With white label, your client hears your company name, calls you when it goes wrong, and it is your name attached to the answer — even though the technology comes from someone else. This is the list you work through first.
Three roles, and you are in the middle
As soon as conversations are processed, everyone shifts one place up the chain. Your end client is the controller. You are the processor. Whoever you hire in is a sub-processor.
That has three concrete consequences. You need a data processing agreement with your end client and one that lines up with your supplier. Your end client has to know who the sub-processors are and be able to object to a new one. And a supplier who can change their chain without notice hands you a problem you did not cause but do have to explain.
So ask your supplier about the whole chain, layer by layer, including the notice period when something changes. The questions that belong with that are in where does your call data live — ask them yourself, before your end client asks you.
The AI disclosure lands in your name
Since 2 August 2026, a caller has to know at the start of the call that they are speaking to AI. That is Article 50 of the AI Act. Exactly who in the chain carries that duty depends on who places the system on the market and who deploys it — but in a white-label setup the caller hears your brand and your client gets the question. Have your own role checked legally; in practice you cannot walk away from it.
In practice that means the disclosure cannot be a checkbox someone switches off by accident. It belongs in the greeting that is written per end client, and you want to be able to hear it per client before the line goes live. If the call is recorded, that belongs in the same sentence. And there always has to be a route to a human: a transfer number your client fills in themselves.
What really has to differ per end client
The dividing line between a usable white-label product and a demo with your logo on it is simple: what can differ per end client?
- Their own opening line, in that company's tone.
- Their own opening hours, and different behaviour outside them.
- Their own way of probing — a law firm asks different things than an installation company.
- Their own transfer number and their own escalation rules.
- Their own knowledge: services, locations, what they specifically do not do.
If a supplier cannot configure that per client, you are selling one product to twenty clients who each needed something slightly different.
The number, and who owns it
Ask early whose account the phone number sits in. If it is in your supplier's account, your end client's number is held hostage there — and so are you. Agree up front: how does a port-out work, who has to sign it, and within what timeframe. A supplier who holds your number holds you.
The same goes for the call data. Does your end client take their transcripts with them when they leave, and in what format?
Who gets the call at two in the morning
This is the question that never comes up in the sales conversation and makes all the difference in practice. Your end client calls you — not your supplier. So put it in writing:
- Within what time your supplier responds, and during which hours.
- Who may change what. Can you adjust an opening time yourself, or does that always go through the supplier?
- Who listens back when a call went badly, and who adjusts the text afterwards.
- What happens during an outage. Does the line fall back to voicemail or a forward automatically, or does nothing ring at all? Ask for the failure mode, not for an uptime figure.
What you have to be able to show
A complaint from an end client is always concrete: an appointment was booked on the wrong day. At that point you want to reach the transcript, the call log and the greeting as it sounded, per client. If you cannot show that, you cannot settle the complaint, and it becomes your word against your client's.
The pricing question you decide beforehand
If you buy per minute, your margin moves with your client's call volume and a busy month costs you money you had not budgeted. That gets genuinely unpleasant if you sold a fixed fee or an unlimited bundle yourself.
A fixed amount per active end client per month makes your margin predictable and lets you set your own price. That is also the model I use myself: a fixed amount per active end client, and what you charge your client is none of my business. Whichever model you pick — pick it before you send out the first quote, not after.
What not to promise
In a market full of over-promising, honest scope is a selling point. Do not promise that the assistant handles emergencies: it should transfer them. Do not promise it replaces a member of staff. And do not promise a fixed processing region you do not have in your own contract — that becomes an unkeepable promise the moment someone digs in.
What I deliver and on what terms is on partners; what the assistant does in practice is on the AI receptionist page. If you are after a custom integration rather than a phone assistant, look at AI & automation.