AI ReceptionistChatbotMissed CallsSmall Business

A chat widget on your site won't stop a missed call in Chilliwack

A website chatbot answers typed questions from people already on your site. It won't pick up the phone. Here's what books the job, and what it costs.

By Reuben Mann8 min readLast updated: 2026-09-25

What an owner searching for a chatbot is usually trying to fix

Somebody searching for a website chatbot at 9pm is not shopping for a widget. They're trying to stop losing the calls and messages that come in while they're on a roof, under a sink, or halfway through a colour appointment with no free hand. A chatbot only catches people already typing on the site. The phone is a different problem, and it's usually the bigger one.

A Langley plumber who searches this term at night is not browsing chat software for fun. He's thinking about the call he missed that afternoon, the one where the customer said they'd try someone else. A chatbot on a website answers people who are already on that website and already typing, which is a narrower group than the people calling the shop directly. Trades and clinics get calls to the number on the truck, the number on the sign, the number a regular has saved from three years ago. The number on the truck, the number on the sign, the number a regular has saved from three years ago. None of that traffic ever touches a chat widget.

OpenAI's news team published a piece on 2026-09-23 about a company called Ringg building phone agents on OpenAI's models that resolve a large share of customer calls without a person picking up. That's a signal about where the technology is heading generally, not a MannVenture number, and we're not going to borrow their percentage and pretend it describes us. What it does confirm is the direction: the phone, not the chat window, is where this kind of automation is proving out first, because the phone is where small businesses bleed work.

So before spending on a chatbot, it's worth asking which channel is losing you business. If it's the phone ringing while your hands are full, a chatbot doesn't touch that at all.

What a chatbot does well, and where it stops

A website chat tool answers a visitor's question in text, on the page they're already on, and some versions can hand off to a booking link. That's real and worth having. It does not ring, does not know who's calling, and does not reach the roughly two-thirds of small-business contact that BC owners tell us still comes in by phone, not by typing.

A website chat tool works the way you'd expect: a visitor types a question, gets an answer pulled from your pages or a script, and maybe clicks through to a form. One widely used chat tool prices its growth plan around $49.17 USD a month billed yearly, starting from 250 conversations, and its AI add-on runs about $32.50 USD a month for 50 AI-handled conversations billed yearly. The plan most small sites buy doesn't add any extra AI allowance on top of that fifty; the free AI allowance is a one-off, not a monthly refill. So the chatbot conversation you're picturing, one that handles a real volume of questions every month, usually means stacking a second paid tier on top of the first.

That's a workable tool for a business whose customers browse before they buy: a med spa comparing treatments, a law office reading through service pages before calling. It does nothing for the caller who never opens the website at all. Nobody calls a roofer's chat widget from a moving truck. They call the number.

We don't sell a chat widget as a separate product because it solves a smaller slice of the actual problem than the name suggests. The AI Website (/ai-websites) answers visitor questions and books work directly on the site for $97 CAD a month, and it's built to be the thing a visitor interacts with once they're already there. If typed inquiries from the site are where you're losing people, that's the right tool. If it's the phone, it isn't.

What answers the phone, and what that costs

A per-call answering service, a receptionist hire, and an AI receptionist are three different cost shapes, not three prices for the same thing. One bills per call on top of a base fee, one is a fixed salary whether the phone rings or not, and one is a flat monthly number that books the job instead of just taking a message.

A per-call human answering service publishes plans like $300 USD a month for 30 calls then $11.50 a call, up through $2,100 USD a month for 300 calls then $8.50 a call, and it charges booking, notifications, recording and routing as separate add-ons on top: $1.50, $0.50, $0.25 and $1.50 a call respectively. Answered by a person is the selling point, and it's a real one for some callers, but the invoice grows with every busy month, which is the month you most need it to hold steady.

Hiring is the other comparison owners run in their heads. BC's own labour-market data puts a receptionist's median wage at $24.00 CAD an hour, up to $30.00 at the high end, which works out to roughly $37,670 CAD a year, and that's before CPP and EI, and before you've covered evenings, weekends or the three days she's sick with the same flu the front desk always catches from customers in January. The job as BC describes it is generally Monday to Friday, which means nights and weekends default straight back to voicemail or your own cell, answered from wherever you happen to be.

An AI receptionist is a third shape entirely: a flat $97 CAD a month for the AI Receptionist (/ai-receptionist), answering every call and booking it, 24/7, on the number the business already has. No per-call meter, no add-on fee for the booking itself, no gap on weekends. That flat number is the whole argument. A per-call answering service punishes a busy month. A human hire is a fixed cost whether the phone rings once or fifty times. This is priced the same in July as it is in December.

The part that matters: booking, not just answering

Taking a message and booking a job read the same in a sales pitch and are not the same product. A message still needs a callback the next morning, by which point the caller has usually already called somebody else. Booking closes the loop on the call itself, and it's the one feature difference worth checking before you sign anything.

Voicemail takes a message nobody returns fast enough. A per-call answering service, at its base tier, mostly takes a message too, unless you're paying the extra $1.50 for booking on top of the per-call rate. A virtual receptionist reading from a script can take a message and still get the trade wrong, sending an HVAC call through as a plumbing one because the script wasn't written for your business specifically.

Booking means the caller hangs up with an appointment on the calendar, not a promise that somebody will call back. For a Kelowna dental office or a Fort Langley HVAC crew, that's the difference between a slot filled tonight and a no-show tomorrow because the patient booked with the clinic that answered on the second ring. It's also the reason a flat-fee system that includes booking is worth more per dollar than a cheaper one that only takes messages, even before the per-call meter starts running.

There's a real caveat here and it's worth stating plainly: some calls should not go through an automated system at all. A complex insurance dispute, a upset customer, a job that needs a judgment call only the owner can make, those are routed to a person or scheduled as a callback. The AI Receptionist is built to close the calls that don't need that judgment call, which in a typical trade or clinic is most of them, and to hand off the rest cleanly rather than pretend to handle something it shouldn't.

Where a chatbot is still the right answer

A business that gets more written questions than phone calls, one whose customers research before they book, is better served starting with the website than the phone. The AI Website (/ai-websites) and the AI Receptionist (/ai-receptionist) solve two different leaks, and paying for the wrong one first wastes a month finding that out.

A wedding and event rental business, a research operations platform, an online fitness coaching practice: these get a real share of their inquiries typed, through a contact form or a chat window, from someone comparing options before they ever pick up the phone. For that kind of business, a chat tool that answers questions on the site and hands off to booking is worth the subscription cost before a phone system is, because most inquiries already arrive typed, not called in.

Most of the trades and clinics we build for don't fit that pattern. A roofer, an electrician, a plumbing crew: the call comes from a truck, a job site, someone standing in their kitchen looking at a leak. Nobody researches a burst pipe. They dial the number on the invoice from the last job. For that caller, the AI Website's chat function is a nice-to-have and the phone is the whole business.

The test is simple: pull up your own call logs against your form submissions for the last three months. If the phone rings more than the form fills, a chatbot is solving the smaller problem first.

Frequently asked questions

Yes, and for a business with both meaningful phone volume and meaningful web inquiries, running both channels covers more ground than either alone. They're separate products at separate monthly prices, so it's worth checking your own call-versus-form volume before buying both.

The system doesn't pretend to be a person, and it doesn't need to hide what it is to book a job correctly. Most callers care whether their appointment gets scheduled, not what answered the phone.

It routes the caller to a person or schedules a callback. The design goal is closing the calls that don't need a judgment call, which covers most routine bookings, while sending the harder ones to someone who can make that call.

Not usually on day one. Owners run it alongside their existing setup for a few weeks to see which calls it handles cleanly before changing staffing plans around it.

See what a buyer finds when they search for you.

Start with the free report card. Your website graded in five areas, in about a minute.