What does an AI receptionist cost?

The benchmark: what a human costs for the same call

The honest comparison is bot against the answering service most plumbing companies already know. Those prices are public too.

Service	Entry plan	Bigger plans
Ruby	$250 a month for 50 receptionist minutes	$395 for 100; $720 for 200; $1,725 for 500
Smith.ai, human	$300 a month for 30 calls, $11.50 a call over	$810 for 90 calls; $2,100 for 300
AnswerConnect	$350 a month for 200 minutes plus $49.99 setup; $2.50 an extra minute	$395 for 300 minutes; $1,625 for 1,200

Do the division on the entry plans. Ruby's works out at $5.00 a receptionist minute and AnswerConnect's at $1.75, against $0.11 to $0.25 for the machine's raw minute. That gap is a person's wage, and for one kind of call it's worth it: the angry caller, the frightened one, the one who needs talking down before they'll give an address. A human beats anything I build at that. The machine wins at the ninth ring at 2 a.m., with no meter running while the caller describes the noise the boiler makes.

So compare prices for the calls you actually miss. The per-call maths and the heat-wave queue problem are in the HVAC comparison; a plumbing phone in a cold snap behaves the same.

Worked example: 120 calls a month, three minutes each

Say your shop takes 120 calls a month and the average one runs three minutes. That's 360 minutes of talk, and here's what each layer costs for it.

Layer one, wired yourself. Say 360 minutes at the $0.11 cheap stack is about $40, and at $0.25 it's $90. Add a number, $1.15 at Twilio or $10 at Retell. Call it $41 to $100 for the month, and every hour of setup is yours.

Layer two. On Goodcall's Starter, say those 120 calls came from 120 different people: 100 are included and the other 20 cost $0.50 each, so $89. If most are repeat customers it's $79 flat. On Smith.ai's AI tier, 25 calls are free and the other 95 are $3.00 each, so $285. Their Pro plan's $150 fee plus $2.00 a call comes to $390 at this volume, so the free tier is cheaper until your call count climbs.

The human benchmark. On Ruby, 360 minutes is past the 200-minute plan, so you're on the 500-minute one at $1,725. On AnswerConnect, 300 minutes are $395 and the other 60 are overage; at the $2.50 a minute their entry plan charges, budget around $545.

Layer three. Mine starts at $4,500 up front, fixed scope, quoted before I touch your phone. The monthly covers the phone line, the minutes and my monitoring, typically several hundred dollars a month at single-shop volumes. In year one it's the most expensive AI line on this page, and I'd rather write that than have you work it out.

If you're a UK sole trader running the same 360 minutes, the raw minutes come to roughly £30 to £70 and Goodcall's $79 is about £60 a month, billed in dollars, before VAT, and before anyone has written a word of your script.

Now look at the spread. Raw minutes, say $41. Goodcall, $79. A build, thousands. The difference isn't minutes. It's the script, the testing, the monitoring, the fallback, and the person who answers when it breaks. You do that work or you pay someone to. That's the whole decision.

One won call-out at HomeAdvisor's average plumber visit of $341 pays for four months of Goodcall's Starter. Mine needs about thirteen to cover the build fee before the monthly is counted. If your tally of missed calls doesn't get near that, buy the cheap one. Run your own numbers before you talk to anyone, including me.

Layer three: what the build fee actually buys

I sell this tier. Here's what the money goes on, then the case against.

• The script, from your real calls. Not a form. I write it from transcripts of the calls your phone already takes, so it asks about the airlock and the stopcock, not your plumbing needs.
• Testing before a customer hears it. I run it against your real calls until it gets the address, the job and the urgency right.
• The fallback route. Every build ships with a route to your existing voicemail or answering service. When the platform goes down, and they do, the phone still rings somewhere.
• Monitoring in the first weeks. The street names it mishears, the callers it confuses, the one question nobody answers. Fixed while it's cheap to fix.
• A person who answers. My number, not a ticket.

None of those need to be me. All of them need to be somebody, and the plan price assumes that somebody is you.

Now the case against. If you run one van, take fewer than 100 different callers a month, are happy to type your own script on a Sunday, and would shrug at a misheard address once a month, buy Goodcall's Starter and keep the $4,500 in the business. If you'd rather people answered and might move to AI later, Smith.ai sells both on the same site, so start on their human plan and cross over later. Neither is a bad decision. Buying my build for a phone that takes ten calls a week would be.

When not to buy any of it, and what to do this week

Three cases where the right price is zero.

• Your phone rings twice a day and you answer both. There's nothing to catch.
• Someone in the office answers and knows the board. They're already better than anything on this page. Talk about after hours, not replacement.
• What you want is booking into Jobber or ServiceTitan. None of these do it, and mine won't either, on purpose. It doesn't know where your vans are, and no phone bot does.

What AI can't do at any price: quote the job, promise a time, decide if the noise is the pump or the pipe, or calm someone who's already shouting. It catches the call and hands it to you. The plumbing is still yours.

Do this this week. For five working days, keep a tally in the van: every call you missed, every voicemail that never called back, and roughly how long each answered call ran. That's your call count and your minutes, the two numbers every price on this page turns on. Put the count into the calculator with your average job value. If it comes out small, buy nothing. If it comes out ugly, pick your layer: Goodcall if you'll do the work, a human service if your callers need soothing, a build if you want it done once and watched. If you want a second pair of eyes on it, book a growth call. Thirty minutes, free, and one of my three answers is none of them.