"Robot receptionist": what people mean and what exists

Type "robot receptionist" into Google and you'll get three very different things: a humanoid on wheels greeting people in a hotel foyer, a phone menu that makes you press 1 for sales, and software that answers your calls in a human-sounding voice. They're not variations on a theme. They're three separate products at three separate prices, and only one of them is worth a small business's attention. Here's how to tell them apart.

Why the phrase is such a mess

"Robot" has meant two things for about sixty years. In science fiction it's a body — arms, a face, something that moves. In engineering it's just automation: a machine doing a job a person used to do. A robot vacuum has no face. A robot arm in a factory has no opinions. So when someone searches "robot receptionist", half of them are picturing a shiny android behind a desk and half are picturing a computer that handles the phone.

The confusion matters because the two have almost nothing in common. One is a hardware purchase with a shipping date. The other is a phone number setting you change in a minute. If you've been put off the idea because you imagined the first, you've been judging the wrong product.

Type one: the physical robot in the lobby

These genuinely exist. You'll have seen them in airport terminals, bank branches in Asia, car showrooms, the occasional hospital atrium. Usually a waist-high unit with a screen on its chest, trundling about, offering directions. SoftBank's Pepper was the famous one, and its production was wound down a few years ago after a run of pilot projects that never quite became normal operations.

They are good at a narrow thing: being interesting. A reception robot draws attention, it makes a brand look forward-thinking, and it gives bored people in a waiting area something to prod. That's a marketing budget line, not a staffing one.

What they're bad at is the actual job of a receptionist. A receptionist's work isn't standing near a door. It's the phone, the diary, knowing that Mrs Patel's boiler has been playing up since March, and chasing the quote that never went out. A robot in your lobby does none of that — and most small businesses don't have a lobby. If you're a plumber, your reception area is the inside of a Transit.

So: real, but rare, expensive relative to the benefit, and aimed at organisations with footfall to impress. Not the thing you need.

Type two: the phone menu — which isn't AI at all

"Thank you for calling. Press 1 for sales. Press 2 for accounts. Press 3 to hear these options again." This is what most people actually hate when they say they hate robot receptionists, and it's worth being precise: there is no intelligence in it whatsoever.

The technical name is IVR — interactive voice response. It's a decision tree someone drew once. The caller's job is to translate what they want into the nearest available number, and if their problem doesn't match any of the options, tough. Older systems added speech recognition, which is where "I'm sorry, I didn't catch that" came from. Same tree, worse input method.

Two things make IVR genuinely annoying. First, it puts the work on the caller — they have to learn your internal structure before they can ask a question. Second, it's usually a queue in a costume. You press four buttons and then wait anyway. For a business with eleven departments it's a reasonable traffic system. For a business with one van, it's a wall between you and work.

If you want to understand how the plumbing of diverts, VoIP and virtual numbers fits together underneath all this, we wrote that up separately in the small business phone system guide.

Type three: a voice that just answers

The thing that actually exists now, and the category we sell into, is none of the above. No body, no menu. The phone rings, something picks up, says your business name, and has a conversation. The caller doesn't press anything. They talk the way they'd talk to a person who'd just answered: "Hiya, my shower's dripping, can someone come out this week?"

Underneath it's three pieces stitched together — speech to text, a language model, text to speech — running fast enough that the pauses feel normal. The interesting part isn't the technology. It's that the caller no longer has to simplify their request. They can ramble, change their mind halfway, give the postcode wrong and correct it. That's the difference between IVR and this, and it's the whole difference.

It's also, honestly, not a person. We say so on the answering service page in those words: the one thing you can't change is that it isn't a person. Everything else — the voice and manner, what it knows about your prices and services, what it does after hours — is a setting.

The three side by side

What "answers the phone" means in practice

Vague promises are easy, so here's the specific scope of ours. It answers in your business's name, 24 hours a day. It takes the job details properly rather than just grabbing a name and number. It can book and quote — but those land in your inbox for approval, and nothing reaches a customer without your yes. You get a WhatsApp update the moment a call ends, plus a recording and a full transcript.

You keep the number on your van. One dial code forwards only the calls you miss, so when you can answer, you answer, exactly as now. Nothing to reprint, nothing to explain to regulars.

And the pricing is flat rather than per call: £29 a month for 100 included minutes, £69 for 300, £115 for 600, then £0.25 a minute beyond that. A busy month costs the same as a quiet one until you run past your minutes. There's a 7-day free trial with 30 call minutes and no card, which is the only sensible way to judge whether the voice suits your callers.

If you want the step-by-step of a single call — greeting to approved quote — that's in what actually happens when an AI receptionist answers.

Will callers mind?

Some will. That's the straight answer, and pretending otherwise would be daft. Plenty of people are fed up with being handled by software, usually because their previous experiences were the press-1 variety or a chatbot that wouldn't let them escape. We've written about that resistance honestly in some customers really don't want to talk to an AI.

What we've found matters most is the comparison the caller is actually making. They're not choosing between your AI and your best self on a quiet Tuesday. They're choosing between your AI and voicemail — or between your AI and ringing the next name on Google. Most people would far rather tell a machine about their leaking shower than talk to a beep.

The greeting carries a lot of weight here, and so does the voice you pick. What callers actually hear goes through the choices, including the greeting trap of trying to pass it off as a human when it isn't.

Where this isn't the right answer

A few situations where you should walk away from the whole category, ours included.

• You already answer nearly every call. If you've got an office manager who picks up on the second ring and you never work evenings, you're paying to fix a problem you don't have.
• Your calls are emotionally heavy. Bereavement services, safeguarding, mental health, anything where the first thirty seconds needs a human being who can go off-script and sit with someone. Don't automate that front door.
• Your business is relationship-deep, not enquiry-led. If your phone mostly rings with existing clients you know by name and your calls are long, conversational and specific, a general-purpose voice will feel thin. Clinics are the interesting middle case — new enquiries suit it, ongoing patient calls often don't. We looked at that split in clinic call answering.
• You want a face in the foyer. Then you genuinely do want a physical robot, or more likely a person. Phone software won't greet anybody.
• You need true complex triage. Deciding whether a fault is dangerous, or whether to send two engineers, is a judgement call. Ours gathers the detail and tells you fast; it doesn't make the decision for you — by design, since nothing goes out without your approval.

How to test one properly

Don't read reviews. Ring it. Thirty minutes of trial minutes is more than enough to learn what you need to know, if you spend them deliberately.

1. Ring it yourself and behave like your worst caller: mumble, change your mind, ask a price question you'd struggle to answer.
2. Ring it from a noisy place — a van with the engine running, a salon with dryers on.
3. Ask something it shouldn't know, and see whether it invents an answer or says it'll have someone come back to you.
4. Check what arrives on your phone afterwards: is the message clear enough to act on without listening to the recording?
5. Get someone who isn't you — a partner, a mate, a regular customer you trust — to ring it cold and tell you honestly how it felt.

If it fails any of those, you've learned something useful for £0 and twenty minutes. If it passes, the only remaining question is cost against the arithmetic you ran above, and we've laid out the full pricing picture in AI receptionist cost UK — including where a traditional human service still works out better.

The phrase "robot receptionist" will probably keep meaning a humanoid for another decade, because that's the picture in everyone's head. The useful version has no body and no buttons. It's just the phone being answered at twenty to ten on a Tuesday night, by something that knows your prices.