AI receptionist for a trade business: what it actually does, and what it should never do
Automations
Somebody rings about a boiler at ten past two. You are on a roof. By the time you are down and have looked at your phone, they have rung two more numbers and one of them answered.
"AI receptionist" is the phrase people have started typing into Google for the fix. It is also a phrase that now covers everything from a robot that answers your calls in a fake American voice to a text message with your name on it. This is what the version we run actually does, what it costs in the market, and the four things it should never be allowed to do.
Two very different things get sold under the same name
The first is a voice bot. It picks up the phone, talks to the caller and tries to take a message or a booking. Some of them are decent now. Most tradesmen I have put one in front of hate the idea of a customer getting a machine on the phone, and I have not seen one yet that handles "it's making a banging noise, can you come today" the way a person would.
The second is what we build, and it does not answer the phone at all. It answers the missed call. Within seconds of a call you could not take, the caller gets a text from you, by name. If they reply, an assistant on WhatsApp takes the details. When the job needs pricing, it hands to you. When you have priced it and the customer says yes, it books them in. The phone still rings you first. The system only exists for the calls you miss.
If you have read what a missed call actually costs you, you already know why the second version matters more than the first. Moneypenny's survey of UK micro businesses found that 69% of callers who reached voicemail hung up without leaving a message. The problem was never that nobody answered. It is that nobody got back to them before they rang the next name.
What it actually does, in order
The text-back. A missed inbound call fires a message within seconds. Ours reads, more or less: sorry I missed your call, tell me what you need and I will come straight back to you. It is sent from your number, in your name. Family, mates and suppliers are tagged before it is switched on so they never get one.
The details. If they reply, the assistant asks what the job is, where it is and when they want it done. Plain questions, one at a time. It stops asking as soon as it has enough.
The photos. For a fuse board, a boiler, a bathroom, anything you would normally drive out to look at, it asks for a couple of photos with a short checklist for that job type. The photos land in the CRM against the contact. For an electrician on a twenty-mile radius, that is a survey visit that no longer happens.
The hand-off. The assistant tells the customer it is passing the job to you, and goes quiet. You see the job, the photos and the whole conversation on your phone, and you send the price.
The booking. When the customer replies with a yes, the assistant finds a slot and books it. They get a confirmation and a reminder the day before. After the job, a review request goes out with your Google link.
The full flow, with the actual messages, is on our automations page. This is what runs for LMB Plumbing & Heating in Pencoed and Renovolt Electrical in Bridgend today.
Why the speed is the whole point
The research on this is old and it has not been beaten. In 2007, InsideSales.com and Dr James Oldroyd, then at MIT, studied three years of data from six companies, more than fifteen thousand leads and a hundred thousand call attempts. The odds of getting hold of a lead if you responded in five minutes rather than thirty dropped a hundredfold. The odds of qualifying it dropped twenty-one times.
Oldroyd went on to co-write the Harvard Business Review piece in 2011 that audited 2,241 companies with a test enquiry. A quarter took more than a day to respond. Nearly a quarter never responded at all. Firms that made contact within an hour were nearly seven times as likely to have a proper conversation with the buyer as firms that waited an hour longer.
Those were sales teams with desks and headsets. A plumber under a floor is not going to beat a five-minute window by hand. The text-back does it in seconds, every time, and that is the entire reason it earns its keep.
What it should never do
Give a price. The assistant gathers the job and the photos. You price it. The moment a bot starts quoting, you get bookings from people who have not agreed a number, and on a one-man operation a cancelled day is the day's money.
Text people who did not ring about a job. Your mum, your mate, the merchant. They are in your phone book, and if you import your contacts into any system they are in that too. Every one of them gets tagged and set to do-not-disturb before anything goes live, and every automation checks for those tags before it sends. It is the first job in the setup, not an afterthought.
Book before a price is agreed. The assistant waits for the customer to say yes to your number. Only then does it book. That timing is the difference between a full diary and a diary full of no-shows.
Fire on calls you made. If the trigger is not limited to inbound calls, the system texts "sorry I missed your call" to a customer who did not answer when you rang them. It looks automated and careless, and it is the fastest way to lose trust in the whole thing.
The rules it has to work inside
Three sets of rules decide how this can be built, and any provider who cannot explain them should not be building it.
WhatsApp. Meta's rules for the WhatsApp Business Platform say that when a customer messages you, a 24-hour customer service window opens. Inside it you can send free-form messages. When it closes, you can only send pre-approved templates, in three categories: utility, marketing and authentication. A missed phone call does not open a window, so the text-back over WhatsApp has to be an approved utility template. That is why it is worded the way it is, and why some of it runs over SMS instead.
PECR. UK marketing texts are governed by the Privacy and Electronic Communications Regulations, enforced by the ICO. The review request is a marketing message. It is allowed to existing customers about similar services under what the ICO calls the soft opt-in, as long as there is a simple way to opt out in every message. So every review text carries a STOP, and a STOP is honoured automatically, across every channel, for good. There is a longer plain-English read of this in the database reactivation post.
Google reviews. Google's Business Profile policy prohibits offering incentives for reviews and prohibits content posted because of one. It allows businesses to "solicit or encourage the posting of content that does represent a genuine experience, without offering incentives to do so or attempting to influence the rating or the contents of the review." So the review request goes to everyone who had a job done, with the same public link, and nobody gets filtered out for being unhappy. If a customer was not happy, the message before the review ask invites them to reply and say so.
What the market charges
If you want a voice bot on a monthly plan, the published prices in the UK sit in a fairly narrow band. IONOS lists its AI receptionist at £39, £69 and £99 a month excluding VAT depending on call volume. Dialzara, a US provider that sells into the UK, lists $29 to $349 a month by minutes used. Several UK answering services do not publish a price at all and route you to a sales call.
What none of those plans include is the bit that took us longest with Renovolt: tagging the phone book, writing the per-job photo checklists, getting the WhatsApp templates approved by Meta, and wiring the hand-off so the bot stops when you start pricing. The software is the cheap part.
Is it worth it for a one-man band
It depends on one number: how many inbound calls you miss in a week. If the answer is two, and you already ring everyone back inside the hour, probably not. If the answer is ten and you are on the tools from eight until five, then by the Moneypenny figure seven of those ten never left a message, and by the Oldroyd figures most of them were gone within the half hour. The text-back alone changes that. The rest of the system is there so the reply that comes back at 2:14pm does not sit unanswered until you are in the van.
What I am not claiming
I have not given a conversion rate for our own clients' setups. We have not run them long enough to publish one honestly.
The "80% of callers hang up" and "tradesmen lose £24,000 a year to missed calls" figures that circulate online could not be traced to any study. They are not in this article.
The legal section is a plain-English read of published guidance and policy, not legal advice.
Sources
Moneypenny, Small Business Call Report. Published 2016. Survey of 300 UK micro businesses (0 to 9 employees) plus call data from 10,000 businesses.
InsideSales.com and Dr James Oldroyd (MIT), Lead Response Management Study, executive summary, 2007. Presented at the MarketingSherpa B2B Demand Generation Summit, 16 October 2007.
James B. Oldroyd, Kristina McElheran and David Elkington, "The Short Life of Online Sales Leads", Harvard Business Review, March 2011.
Meta for Developers, WhatsApp Business Platform documentation, "Send messages", section on the customer service window and message templates.
Information Commissioner's Office, Guide to PECR: Electronic mail marketing.
Google Business Profile Help, "Prohibited and restricted content", reviews section. Quoted wording is Google's.
IONOS UK, AI Receptionist product page, and Dialzara pricing page, both read 3 September 2026. Prices as published on that date.
Do the maths
Matthew Betts
