All posts

Field notes

10 October 202614 min readBy Team Klaudius

Klaudius can now pitch personalised AI receptionists at scale: every prospect rings one number and hears their own

Just like the websites, every business Klaudius pitches can now get a working AI receptionist of its own before they've spoken to you. Your pitch to the business can include a number for them to call; the owner rings it and hears a personalised receptionist answering for their business, already knowing it from the research Klaudius did. One number covers every prospect you'll ever pitch, ten or five hundred, for a dollar or two a month in total, not per prospect. Once they sign, it answers their real calls around the clock and texts the owner every job.

You already know how Klaudius pitches websites: it builds a finished site for every business it contacts and sends them the link to their own. Klaudius can now do the same with an AI receptionist. Turn it on, and every business your outreach touches gets a working receptionist, personalised to them, before they've spoken to you.

Your pitch to the business contains a number for them to call. The owner rings it and hears their own personalised receptionist answer: it greets their business by name, knows their trade, and answers questions about their services the way it would for one of their customers. They're not being told about a product. They're hearing their own phone being answered.

And you don't build any of them. It all happens on its own, inside the same autonomous run that finds, builds and pitches. One demo number covers your whole operation: not one per prospect, one in total. Every outreach run writes the business's profile from the research it already gathered, registers it on your number, and adds a line to the pitch inviting the owner to ring it. There's no step per prospect, on any of them.

Ten prospects or five hundred: same number, same dollar or two a month in total, not per prospect. A fully working, personalised demo for every business you pitch, built without you lifting a finger.

Here's how the pitch works, what the client gets once they sign, what it costs to run, and how to sell it.

New here? The short version: Klaudius is an AI pipeline you own that finds local businesses that need a website, builds each one a bespoke site, and pitches it to them — you charge the businesses. You don't need to code. One plan with everything in it, monthly or annual, with a 7-day money-back guarantee. See how it works → — or keep reading for the full detail.

How the pitch works: one number, a receptionist tailored to every business you pitch

  1. Klaudius researches the business. For a website pitch, the research already exists from the build: services, service area, opening hours, reviews.
  2. It registers the business on your demo line before the pitch goes out, under the phone numbers it has for them.
  3. The pitch includes the number. The message invites the owner to ring it and hear the receptionist you'd set up for them.
  4. The owner rings, and it answers as theirs. It recognises the number they're calling from, greets their business by name and talks about their work. If they ring from a phone it doesn't know, it asks for the business name and finds them anyway.
  5. It tidies up after itself. Registrations expire on their own once the pitch window passes, so there's nothing to manage.

What the pitch costs you. The one number is about $1–2 a month in total, however many businesses you pitch, and you only pay for the minutes prospects actually spend on the line, a few cents each.

Or sell the receptionist on its own

Plenty of businesses already have a website they're happy with. They still miss calls. So Klaudius can also pitch the receptionist by itself: it looks for businesses likely to be losing work to missed calls, researches each one, registers it on your demo line and sends a pitch for the receptionist alone, with no website build. Ask Klaudius to switch to receptionist-only outreach, or to keep pitching both.

What it fixes for the business

Think about who these businesses actually are. A plumber with both hands inside a boiler. A roofer three storeys up. A lash technician mid-appointment with a client in the chair. A one-van electrician driving between jobs. For hours of every working day, these people physically cannot answer the phone — and the phone is where their work comes from.

And a caller who gets voicemail doesn't leave a message. They're standing in a flooding kitchen with a list of search results open. They hang up and ring the next plumber on the list. The first business never knows the call happened, never knows what it was worth, and never gets a second chance at it. A single missed emergency call-out can be a few hundred dollars of work handed straight to a competitor.

Businesses know this, which is why call-answering is already a paid industry — human answering bureaus charging from around a hundred a month into the high hundreds. Your clients are exactly the market those services chase.

What the client gets once they sign

The design principle that makes this sellable: the AI never replaces the owner. It catches what the owner drops.

Each client gets a dedicated phone number, and their existing phone is set to conditional forwarding — three dial codes typed once, which Klaudius generates as a simple handover sheet for their exact carrier. Their phone still rings first, every time, exactly as before. Only when they're already on a call, don't pick up, or have no signal does the call divert — and instead of voicemail, it reaches their receptionist:

  1. It answers instantly — no dead air, no hold music — introduces itself as an AI assistant for the business by name, and explains the team couldn't get to the phone.
  2. It handles the call like a good receptionist would. It knows the business's services, opening hours and service area, asks what the caller needs, and collects a name, a phone number confirmed digit by digit, and enough detail about the job for the owner to quote or prioritise.
  3. Emergencies and "can I speak to someone" go straight through. If the caller describes something urgent — water pouring through a ceiling, a suspected gas leak — or simply asks for a human at any point, the AI transfers the call to the owner's mobile immediately, no questions first. For trades with a national emergency line, it says that too: a gas company's receptionist tells a caller who smells gas to ring the gas emergency number and open the windows, alongside putting them through.
  4. The owner gets a text while the caller is still on the line. Name, number, one-line job summary — sent the moment the details are collected, not after the call. The owner glances at their phone under the sink, sees "burst pipe, Kings Heath, can pay today", and decides in five seconds whether to call back now or after the job.
  5. It closes honestly. The caller is told when to expect a callback — a realistic window anchored to the business's actual opening hours — and nothing more. It never books jobs, never quotes prices it wasn't given, never promises the business will take the work.
A real call to a client's receptionist, with the owner's text arriving mid-call. Bridgeworth Plumbing is a test business, and the phone numbers are changed. Sound on.

The whole thing runs 24/7. At 2am on a Sunday, the receptionist answers exactly as brightly as at 10am on a Tuesday — which is precisely when a burst-pipe call is worth the most.

The design decisions that matter

We spent weeks running this against real phone calls before shipping it, and the details that survived that testing are the reason it's trustworthy enough to put in front of a paying business.

It says it's an AI, first sentence, every call. Not buried, not optional. AI-disclosure rules are tightening in several markets, and beyond compliance it simply works better: "I'm an AI assistant — I answer when the team can't get to the phone" sets honest expectations, and callers get on with describing the job. It leads with the truth on every call.

It cannot fabricate. The hard rule set: never claim an action happened unless it did. It will never tell a caller "Dave's on his way" — it doesn't know that. A transfer attempt is only ever "trying to connect you now". "I've sent your details to the team" is only said after the text has actually gone. If something fails, it says what it couldn't do and takes a message instead. Dependable, honest behaviour is the whole product, and it's what owners come to trust.

The owner's text always carries the caller's real number. The number that arrives in the notification is taken from the phone network's caller ID — the same mechanism that puts a number on your missed-call screen — not from the AI's transcription of the conversation. Voice models can mishear or mangle a dictated digit; caller ID can't. The spoken digit-by-digit read-back still happens on the call, because it confirms the caller's intent — but the number the owner rings back is authoritative. If a caller asks to be called back on a different number, that's added separately and clearly.

Written notes, not recordings. The receptionist keeps a transcript so the owner has an accurate note of every call; it doesn't store audio recordings. Callers are told about the written note in the greeting. Simpler, lighter, and a far easier privacy story for your client to stand behind.

Every prompt is generated per business, per country. Phone numbers are read back in the local convention. The emergency guidance matches the trade and the country. The greeting carries the client's business name and the callback promise matches their actual hours — including what "tomorrow" means on a Friday afternoon. None of it is a shared script with a name slotted in.

It speaks the client's language. The prompt, the voice and the speech recognition are all set for the client's market, so a caller in Milan is answered in Italian and a caller in Rotterdam in Dutch, in a voice made for that language. The demo line works the same way, in the language you work in.

And because we know operators will ask: yes, it held up under hostile testing. We ran real calls from silent callers, abusive openings, callers who interrupt, callers asking it to pretend to be human, callers fishing for prices it was never given — plus automated call batteries against every rule above. The behaviours in this post aren't aspirations; they're what the test transcripts show.

What it actually costs to run

This is where it gets good. The receptionist runs on a voice platform account of your own — the same pattern as everything else in Klaudius: your infrastructure, your keys, no middleman metering you.

  • The AI receptionist itself: about $0.05–0.06 per minute of answered call, everything included.
  • The client's dedicated number: roughly $1–2 a month.
  • Each owner-notification text: a few cents.

Run the maths on a realistic client — say 30 forwarded calls a month, a couple of minutes each: about $3–4 of call time, a couple of dollars of texts and number rental. Call it $5–6 a month, all-in. Even a genuinely busy trade missing five calls a day only reaches about $20–25 a month. Our own test bill for an entire month of intensive engineering — dozens of live calls, automated test batteries, the lot — was under ten dollars.

Against what this service sells for, the margin does not need explaining. Operators charge in the low hundreds a month, and almost all of that is margin — recurring, arriving whether or not you sold anything new that month. Ten clients on the receptionist is a serious monthly line of its own. And costs are bounded by design: every call is hard-capped at ten minutes, the account is prepaid — the worst case is the balance you've chosen to load, never an open-ended bill — and a quick daily check warns you if usage passes the monthly minute cap you set.

What setting it up looks like

First, your demo line, once. Ask Klaudius to set up your AI receptionist demo line, and it takes it from there. If you don't have an account with the voice platform the receptionist runs on yet, it walks you through opening one; new accounts go through a one-time verification before you can add credit, which can take a day or two, and a few dollars of credit covers your first tests. Then it buys the number, builds the demo receptionist in your language, and connects the lookup that recognises each business you pitch. It registers a sample business against your own mobile, so you can ring the line and hear it answer as that business. From then on, every pitch registers its business on the line automatically.

Then, each client who signs. Ask Klaudius to set up a receptionist for them. It already knows the business from its research — services, prices, opening hours, service area, reviews — so it only asks you what research can't answer: the owner's mobile for texts and transfers, and a quick check of what counts as an emergency for this trade, from a list it suggests. Then it does the rest itself: provisions the number, builds the per-business prompt, wires the transfer and owner-notification text, applies the safety rails above, and runs its checks. You place one test call, hear it answer, and hand the client a one-page sheet with the three forwarding codes for their carrier.

The client's side of the install is typing three short codes into their own phone — the sheet shows exactly which, with a check code to confirm each one took. No app, no new hardware, no change to their number. Cancelling is the same codes backwards, which is worth saying in the pitch: trying it costs them nothing and changes nothing about their phone.

It's region-agnostic like the rest of Klaudius. US numbers activate instantly (US carriers do require a one-time sender registration before the owner-notification texts flow — a form plus a few days of review, done once with your first install); the UK and some other countries ask for identity documents before a number goes live. Either way Klaudius knows which paperwork your case needs and walks you through it once, and it's reused for every client after that.

Selling it to the clients you already have

Cold prospects get the demo line. The clients who already bought a site from you are warmer still: you built their website, so you're already the person who "does their internet". The pitch is one message:

"Quick one — when you're on a job and can't pick up, what happens to the call? I've set up a 24/7 receptionist service for a few of my clients: if you don't answer, an assistant picks up, takes the caller's name, number and what they need, and texts it straight to you. Your phone works exactly as it does now. Fixed monthly fee, cancel anytime — want me to set it up so you can hear it?"

And "hear it" is the close, again: give them the demo number, they ring it, and it answers as their business.

It also stacks with everything else. The booking-mode clinics and salons — businesses that live and die by the phone and physically can't answer mid-treatment — are among the strongest receptionist prospects in the pipeline. And for lapsed leads who never bought the site, "I also set up 24/7 phone answering" is a fresh reason to reopen the conversation.

The bigger picture

Klaudius started as a machine for selling builds, and plenty of operators already make those recurring with care plans and retainers on top of the site. The receptionist is recurring by nature, because the value itself is monthly: real work, delivered by AI around the clock, at a cost base that rounds to a takeaway coffee — and unlike a rankings retainer, the client can hear it working.

It's part of the standard pipeline. New subscribers get it out of the box, and if you're already running Klaudius, the app picks it up on its own: ask Klaudius to set up your demo line. On the earlier one-time licence, ask your AI agent to update Klaudius first.

Not an operator yet? Klaudius is one plan with everything in it — the pipeline that finds the businesses, builds the sites and pitches them, plus the AI receptionist in this post, the booking systems, the client CMS and every new capability as it ships. Monthly or annual, with a 7-day money-back guarantee. Start here →