Is an AI receptionist any good for a small plumbing or electrical firm?

Why this trade misses calls in the first place

You're under a sink. You're in a loft. You're holding a board cover with one hand.

That's the whole explanation, and it's not a management failing. A two-van firm has no one at a desk, so the phone is answered when it's convenient to answer it — which is not when it rings. Domestic callers with a problem don't leave voicemails, they ring the next name on the list, and by the time you're back in the van at four the job is gone.

The seasonal version is worse. First cold snap, everyone's boiler fails at once, and the week you get the most calls is the week you have the least capacity to answer them.

What it genuinely does well

Answers within two rings, at any hour. Including seven on a Sunday evening, which for domestic work is a real chunk of enquiry volume.

Establishes what the job actually is. Not just a name and a number — the fault, the property type, whether it's a landlord or an owner, whether there's water coming through a ceiling right now. You ring back knowing whether it's worth ringing back.

Checks coverage. If they're forty minutes outside your patch, that's found out in the first thirty seconds rather than after a callback and a site visit.

Sends you a usable summary. A short written brief on your phone the moment the call ends, rather than a voicemail you'll listen to twice.

Texts back missed calls automatically. The simplest and cheapest thing on this list. An automatic text within a minute catches a real share of callers before they dial the next number, and it costs almost nothing.

Books the straightforward work. A boiler service, a routine callout, an EICR quote visit — these can go into the diary against your real availability without you touching them.

What it can't do, stated plainly

This is the section most vendors leave out, so here it is.

It cannot price a job. Not a rewire, not a boiler swap, not a bathroom. Variables you can't see over the phone decide the number, and a figure quoted by a machine becomes an expectation you'll have to argue with on site. If a vendor tells you their agent can quote your work, they haven't done your work.

It cannot diagnose a fault. It can capture symptoms. It cannot tell someone whether their boiler is safe to run.

It should not be handling a gas or electrical emergency alone. A caller who can smell gas needs to be told to turn off at the meter, ventilate, get out and ring the emergency service number — and then to reach a human. That path must be configured explicitly and tested. Anyone selling you a voice agent who hasn't asked how you want emergencies handled is selling you a liability.

It handles complaints badly. A customer whose repair failed does not want a pleasant automated conversation. Escalate immediately, every time.

It will occasionally get something wrong. Every system does. The questions that matter are whether it escalates cleanly to a human, whether you get a recording and transcript, and whether the vendor will talk about it honestly. If they won't, they've never run one under load.

Will customers be annoyed?

Some will notice. Fewer will mind than you expect, and the comparison isn't "AI versus a receptionist" — it's "AI versus the phone ringing out".

Three configuration choices make almost all the difference.

Keep it short. Four or five questions, then done. Long automated calls are what people actually resent.

Don't pretend it's a person. A brief, matter-of-fact framing — an assistant taking details so the engineer can call back — is better received than an agent that dodges the question when asked.

Escalate fast when it's not going well. If the caller is distressed, confused or repeating themselves, get them to a human. A system that stubbornly keeps trying is where the bad stories come from.

What it's actually worth

Do this with your own numbers, not an industry statistic. The percentages circulating in this market are mostly unsourced and you can get the real figure in ten minutes.

Ask your phone provider for last month's unanswered inbound calls. Halve it, because not every missed call is a customer. Apply your normal conversion rate from enquiry to job. Multiply by your average job value — and be careful here, because a firm doing £120 callouts and £9,000 heat pump installs has two completely different answers depending on which calls are being missed.

That last point is the one worth sitting with. If the calls ringing out are mostly small callouts, the case is decent. If even one heat pump or full rewire enquiry a quarter is going unanswered, the arithmetic stops being close.

If your problem is further down the line — quoted work that never gets booked in — read why quoted jobs never get booked in. For the wider order of operations, see what a small trade business should realistically automate first.

Where OrderWolf fits

We don't hand over an agent and a login. We configure it on your services, your area and your qualifying questions, write the escalation rules including how emergencies are routed, connect it to your diary and your CRM so every call becomes a tracked enquiry, and then we run it and tune it. If a call goes badly, that's our problem to fix, not another job on your list.

With GEV, our Irish flagship client, surveys booked rose 8% and the B2C close rate moved from 45% to 70% across a three-month deployment. Results from one installer — yours will depend on your call volume, your job mix and where you're starting from.