Will your customers accept an AI answering the phone?

This is the first question almost everyone asks, and it deserves a better answer than “customers love it!” — which is what you will get from most vendors, including ones with no evidence either way.

What the question is really comparing

The instinct is to compare an AI answering against you answering. On that comparison the AI loses, and you should keep answering.

But that is not the choice. The realistic comparison is against what happens today when you cannot pick up — a phone that rings out, or a voicemail nobody listens to. Against that, most customers are markedly happier to talk to something that answers.

The exception matters though: if you currently answer nearly every call personally and quickly, an assistant is a downgrade for your customers and you should not buy one. Volume is what changes the answer.

What actually determines whether people mind

In practice it is not “AI vs human”. It is four much more specific things:

Did it answer the question? Almost everything else is forgiven if the caller got what they rang for. Almost nothing is forgiven if they did not.

Was there an awkward pause? The gap before it starts speaking is the single biggest driver of whether a call feels natural. A long pause reads as a bad line, and people start talking over it.

Was it in their language? Someone who has to switch to their second language to be understood has a worse experience regardless of what is on the other end.

Could they get to a person? The thing that makes people angry is being trapped. An assistant that recognises it is out of its depth and gets you to a human is fine. One that keeps confidently going is the experience everyone is actually afraid of.

Should you tell them?

Yes, and not for ethical reasons alone — it works better. Callers who know they are talking to an assistant ask more directly and get better answers. Callers who suspect it and are unsure spend the first thirty seconds testing it.

Most people work it out in a sentence or two anyway. Pretending is a losing position that costs you the goodwill of having been straight.

Where it genuinely goes badly

Three situations where an assistant is the wrong answer and no vendor should pretend otherwise:

• An upset customer. Someone with a complaint needs a person with authority.
• A negotiation. Anything where the price or the terms are being decided.
• An unusual case that needs judgement. The one that does not fit the pattern.

The useful design question is not whether these happen — they will — but whether the assistant recognises them and hands over, or improvises.

Test it on your own customers

Do not take our word for any of this, and do not take a competitor’s testimonial page either. Two things to do before committing:

1. Call the assistant yourself and try to break it the way an awkward customer would — talk over it, change your mind, ask something off-topic, give a phone number, mix languages.
2. Ask five regulars. Not “would you mind talking to an AI” — that gets you the polite answer. Ask instead: “when you called last month and I couldn’t pick up, what did you do?” The answer to that is your real business case.

⚠️ Worth knowing before you test: you get an email when something needs your attention. There is no WhatsApp alert to you, no SMS and no push notification, so “it escalates to me” means it reaches your inbox, not your phone screen.