Blog · Trust and etiquette
Should an AI say it's an AI on a phone call?
Yes. It's the decent thing to do, and it also happens to be the thing that gets you the table.
By the CosVoice team · · 5 minute read
Yes, and it should never claim to be a person
An AI that makes or answers a phone call should say what it is. At the very least it should introduce itself as an assistant, tell the truth the moment anyone asks, and never claim to be a human being or to be you. We don't think it's a close call.
There are three reasons and they stack. It works better, because people help a caller they understand. It's fairer to the person on the other end, who is giving up a minute of a busy day. And the rules in the US are moving that way regardless. Call it AI disclosure if you like the formal name. Most people would just call it being straight with someone.
People help more when they know what they're talking to
Picture the host at a busy restaurant. The voice on the phone is slightly too even, pauses a beat longer than a person would, and doesn't laugh at the joke about Friday nights. If the caller has claimed to be Graham, the host now has a puzzle to solve. Is this a scam? A prank? Somebody reading from a script? Suspicion slows everything down, and suspicious people say no.
Now run it the honest way. "Hi, this is Lloyd, Graham's assistant, I'm calling to book a table for four on Friday." The host knows the deal inside one sentence. There's a real customer behind the call, there's a reason the voice sounds the way it does, and they can get on with the only thing they care about, which is whether there's a table at 7:30.
There's a bonus too. Staff who know they're speaking to an assistant tend to speak plainly. They spell the name, repeat the time, mention the deposit policy. That's the material you want in the written summary afterwards.
The ethical case is short
A phone call is a small act of trust. The person who answers decides how much time, warmth and information to give, and they decide it based on who they think is calling. Let a machine pass as a person and you've taken that decision away from them.
It matters more as the stakes rise. Someone who believes they're talking to the homeowner might read out a gate code. Someone who believes they're talking to the boss might agree to something they'd normally want in writing. Saying what you are keeps everybody's guard at the right height.
And think about being on the receiving end. Would you want to find out afterwards?
Where the rules in the US are heading
General information only, not legal advice. Rules about automated and artificial voices on the phone differ from state to state and from country to country, they're changing quickly, and you should check what applies where you are and where the people you call are.
The direction is clear enough, though. Lawmakers and regulators are paying more attention to artificial voices, not less, and the thread running through it is that people should be told when they're talking to a machine. Some places expect that to be said at the start of the call and not only when somebody asks.
So even with the ethics set aside, a phone line designed to hide what it is has been designed for a world that's closing. Our guide, can an AI legally make phone calls for you?, walks through the wider territory.
How our line introduces itself
On a call it places for you, the line opens with one sentence: who is calling, and on whose behalf. It introduces itself as your assistant. It books under your name, the way a human assistant would, but it never claims to be you and it never claims to be a person.
If anyone asks whether they're talking to an AI, it says yes. Whoever asks, however they phrase it.
Calls coming in work the same way. The caller reaches your assistant, which finds out who they are and what they need, and tells the truth about itself when asked. The rules it keeps are listed on the trust page.
You can check all this in under a minute. Our own number, (949) 368-9902, is answered by an AI agent called Lilly. Ring it and ask her straight out.
What good disclosure sounds like
If you're weighing up any AI phone product, ours or somebody else's, listen to the first ten seconds of a call and check for these.
- It says whose assistant it is. A name and a reason for calling, in one sentence.
- It answers the AI question with a yes. Not a joke, not a change of subject.
- It never borrows your identity. Booking under your name is fine. Speaking as you is not.
- It doesn't make a speech. One clear line, then on with the call. A long disclaimer wastes the time of the very person you're trying to be fair to.
But won't people just hang up?
Some will. A few businesses have a policy of not dealing with automated callers, and some people simply don't like it. That's their right. When it happens on our line, the call comes back to you with an honest status saying they declined, and you ring them yourself.
In fairness to the hang-up crowd, there are calls where an assistant shouldn't be on the line at all, disclosed or not. Condolences. A complaint you feel strongly about. Anything where the other person deserves your own voice. A human call is simply better there, and we've written about where that line falls in is it rude to have an AI call a business for you?
For everything else, we'd rather lose the call than win it by lying. A booking made under false pretences is fragile anyway. The first time the restaurant rings back and an assistant answers, the story falls apart, and it's your name on the reservation.
Common questions
Does an AI have to identify itself on a phone call?
It depends on where you and the other person are, and the rules are changing, so check what applies locally. Wherever you are, the safe choice and the decent choice turn out to be the same one: say what it is.
Will a business refuse to talk to an AI assistant?
Some do, and they're entitled to. When that happens on our line the call ends politely and comes back to you marked as declined, so you know to ring them yourself.
Does the assistant pretend to be me when it books?
No. It books under your name, which is what the business needs, but it introduces itself as your assistant. It never claims to be you.
Can people tell it's an AI without being told?
Sometimes, often not. That's the reason disclosure matters. A voice this natural means you can't rely on the other person working it out for themselves.
Read next: Seven things an AI should never do on a call · How businesses react when an AI calls · The rules our line keeps
Hear it for yourself.
Call our own AI agent and ask her anything, or get a number in your area code in about a minute.