What actually happens when an AI receptionist answers your phone

"AI receptionist" sounds either magic or terrifying, depending on your mood. Neither is right. It's a fairly ordinary chain of events: a phone rings, something answers, notes are taken, and you get a message. This post walks that chain end to end — what the caller hears, what lands on your phone, what happens when it goes wrong, and the sorts of businesses it doesn't suit.

The short version of how it works

Before the detail, here's the whole loop in one breath. Someone rings your number. You don't pick up — you're up a ladder, or it's 9pm. The call forwards to your AI receptionist, which answers in your business's name, has a normal conversation, and gets the job details. The call is recorded and transcribed. A draft quote or booking is created and held in your approvals inbox. You get a WhatsApp message. You tap approve. The caller gets a confirmation.

Nothing reaches the customer until you say so. That's the part most people don't expect, and it's the part that makes the rest of it usable.

Step 1: the call arrives — and your number stays your number

You don't hand out a new number. Your existing number stays on the van, the cards, your Google listing, everything. One dial code on your mobile forwards only the calls you miss to your AI number: *61your-AI-number#.

So the order of events is: your phone rings first. If you answer, nothing else happens — business as usual, the AI never enters the picture. If you're busy, on another call, or it's outside hours, the call rolls over instead of hitting voicemail. Your customer doesn't know or care which of you picked up; they just know someone did.

That's worth sitting with, because it's the bit that changes your day. You stop feeling obliged to answer with a wrench in your hand. The call is caught either way.

Step 2: what the caller actually hears

A voice answers in your business's name. Not "you've reached an automated assistant" — it greets them the way a receptionist would, using what your business actually does.

Where does it get that from? Your website. During setup you hand over your URL, and RedAgents reads the site to learn what you do, who you serve, and how you talk about it. It builds a call script and a business profile from that. If your site says you cover Coventry and Warwick and don't do commercial gas work, that's what the receptionist knows.

The conversation itself is a conversation. The caller talks, the AI responds, asks follow-ups, clarifies. It's not a phone menu — there's no pressing 1 for bookings. It's aiming to get the same things a good human receptionist gets:

• Who's calling and how to reach them
• What the job is, in the caller's own words
• Where it is
• When they need it, and whether it's urgent
• Anything that decides whether you'd even take the work

Some callers will clock that it's an AI. Most people are fairly relaxed about it these days, especially at half nine at night when the alternative was a beep and a voicemail box. The measure that matters isn't whether they notice — it's whether they got a straight answer and a call back.

Step 3: the call is logged, in full

Every call is recorded and transcribed into your dashboard. Not a summary you have to trust — the actual words, plus the audio if you'd rather listen.

This matters more than it sounds. It's how you check that the AI understood the job. It's how you settle a "but I told them Thursday" conversation three weeks later. And in the first week or two, it's how you build confidence: read a few transcripts, hear how it handles people, decide for yourself.

Contrast that with what you have now. A missed call is a number in your log and nothing else. You don't know if it was a boiler swap or a bloke selling SEO. You can't tell what you lost, which is exactly why missed calls are so easy to shrug off — we've done the arithmetic on that in what missed calls really cost a UK trades business.

Step 4: a draft lands in your approvals inbox

Here's where RedAgents deliberately stops short. Off the back of the call it drafts the next action — a quote, a booking, a follow-up — and puts it in your approvals inbox. It does not send it.

An entry looks like: "QUOTE — Boiler service for Sarah M., £145, Thursday 2pm." Or "BOOKING — Confirm callout for 14 Elm Road, tomorrow 9–11am." Three buttons: approve, edit, reject.

So the AI is doing the typing, the listening and the remembering. You're doing the deciding. That split is the whole design. Plenty of tools will happily act on your behalf and hope for the best; this one holds the pen until you nod.

Step 5: WhatsApp tells you, seconds later

You don't have to check a dashboard to know something happened. A message hits your phone within seconds of the call ending: who rang, what they wanted, and what needs your yes.

It reads roughly like this: "New call from 07700 900123 just now. Booking request: boiler callout, Thursday 9–11am, £85. Approve?" You tap approve. It replies: "Done — I'll text Sarah to confirm."

Total time from you, standing in a loft: about four seconds. That's a real change to how the admin feels. The 9pm shift of returning calls and writing up quotes turns into a handful of taps done between jobs.

Step 6: the caller gets their confirmation

Once you approve, the customer gets the text. From their side the experience was: rang a plumber at 8:40pm, spoke to someone, got a confirmation with a time and a price. They have no idea you were watching telly.

"But what if it gets something wrong?"

This is the first question everyone asks, and it deserves a straight answer rather than reassurance.

It will occasionally get something wrong. It might mishear a street name, misjudge urgency, or quote from your website pricing when the job clearly needs a site visit. Any receptionist — human or otherwise — does this in their first month.

What limits the damage is the approval step. The wrong price doesn't reach the customer, because nothing reaches the customer without your say-so. You see "£145" on your phone, you know it's a £280 job, you edit it or reject it. The mistake dies in your inbox.

And because every call has a transcript and a recording, you can always see exactly what was said rather than guessing. If the caller gave a different postcode to the one in the draft, that's checkable in about ten seconds.

The failure mode that actually costs you

Worth naming: the risk isn't a wrong quote sitting in your inbox. It's an approval sitting there for two days because you didn't look. The AI holds the draft indefinitely — that's the point — but the customer is waiting. If you're the sort who lets WhatsApp messages pile up unread, set a habit: clear the approvals inbox at lunch and at the end of the day.

What happens on a complicated call

Not every call is "boiler's broken, when can you come". Some are messy.

On the harder calls, the honest framing is that the AI is a very good note-taker with a friendly voice, not a closer. It gets you the name, the number, the problem and the urgency — which is precisely what a voicemail beep fails to do — and then hands the decision to you.

Where this isn't the right fit

Some businesses shouldn't buy this, and it's better you know now.

• You want long consultative sales calls handled end to end. If your work is won over a 25-minute conversation where objections get handled and trust gets built, an AI receptionist won't do that job. It'll take the enquiry and get you on the phone faster. It won't replace you on the call.
• Your calls are highly regulated or clinical. If a call involves triage, safeguarding, or advice with legal weight, you need a trained human on the line, not a draft in an inbox afterwards.
• You genuinely answer every call already. If you've got someone on reception all day and after-hours cover, the gap you're paying to close may not exist. Look at your missed-call log before you look at plans.
• You won't check the approvals inbox. The whole model depends on you tapping approve. If you know you won't, a traditional answering service with human operators may suit you better — we compared the options in AI receptionist vs answering service vs voicemail.
• Your customers ring about deeply bespoke work with no reference points. If nothing on your website tells the AI what you charge or what you do, it has less to work with. It'll still take a message well. It won't quote sensibly.

How long the setup actually takes

Four steps, and you can watch them happen. You share your website. RedAgents reads it and writes a tailored call script and business profile. A UK business number is provisioned and wired up automatically. Then you set your phone to forward missed calls with the dial code.

No setup call, no card needed for the trial — a 30-minute, 7-day free trial, no contracts, cancel any time. The realistic first-day job isn't the setup, it's the listening: ring your own number, hear how it answers, read the transcript afterwards, and tweak from there.

That's the whole loop. A caller rings, something answers in your name, the job gets written down properly, and you decide what happens next from your phone in four seconds. The AI does the part you can't do from up a ladder. You keep the part that has your name on it.