Smith.ai vs Ruby vs building your own.

We build AI receptionists, so treat this as an interested party writing about its competitors. The way to make that useful rather than worthless is to be specific about which businesses should buy something else, and why. That is most of what follows.

They are three different models, not three brands

The confusion in this category comes from all three being described as “virtual receptionists”. Underneath, they differ in the two things that actually decide your cost and your outcome: who answers the call, and what you are billed for.

Ruby

Ruby is a human answering service. Real receptionists take your calls, and the published pricing is per minute, with plans starting around $245 per month for a bundle of minutes and rising through higher tiers as volume grows.

The per-minute model is the thing to understand. It means a long call costs more than a short one, so your bill moves with how talkative your callers are rather than how many of them there were. If your calls are short and numerous, that works in your favour. If they are long and involved, it does not, and monthly cost becomes hard to predict.

Where Ruby wins: when the call genuinely needs a person. Complaints, distressed callers, sensitive or high-value conversations, anything requiring improvisation. No AI available today handles those as well as a competent human, and pretending otherwise would be the sort of claim that makes this whole page untrustworthy.

Smith.ai

Smith.ai is a hybrid: AI handling calls with a network of live agents behind it, which the company states are North America based. Published pricing is per call rather than per minute — around $195 per month for a small bundle of calls with a per-call rate beyond it, an AI-only tier at a lower monthly figure, and larger plans running into several hundred dollars a month.

Per-call pricing behaves differently. Cost is predictable regardless of how long anyone talks, which suits businesses with involved conversations. The trade is that short calls cost the same as long ones, so a high volume of thirty-second enquiries is expensive per unit of value.

Where Smith.ai wins: when you want something that works out of the box with meaningful capability — qualification and CRM updating rather than message taking — without a build project. It is a genuinely capable product and for a lot of businesses it is the right purchase.

Where both of them stop

Ruby and Smith.ai are productised services, and that is their strength — you start this week. It is also where they end. Both configure within their own model, and there are three points where a growing business hits that ceiling.

The call ends in a notification, not in your systems. This is the big one. A message telling you someone rang is not the same as an appointment sitting in the right diary, against travel time and territory rules, with the qualifying answers already written to the CRM record. The first still leaves the work with you. The second is the work, finished.

Your qualification rules only bend so far. A product configures what its designers anticipated. If your criteria involve roof orientation and shading, or whether a caller owns the property, or which of six branches covers a postcode, you are asking a template to do something it was not shaped for.

The pricing works against you exactly when you succeed. Per-call and per-minute billing scales linearly and never stops. Twice the calls is twice the bill, permanently. A build costs more to start and then flattens — the monthly barely moves whether you take fifty calls or five hundred.

What a built agent does instead

Setup starts at £1,500 with a managed service from £195 per month, plus usage. What that buys is not a better voice — it is what happens around the call.

• Books into a live diary against real constraints. Travel time, skill, territory, branch. Not a request for someone to action later.
• Writes to your CRM. The record, summary, transcript and next step, in the fields you actually use.
• Asks your questions. The qualification script is built from your criteria, however specific, and anything failing them is held rather than booked.
• Escalates on your rules. Urgent, high-value or upset callers reach a person with the context already gathered.
• Answers everything at once. A surge is handled in parallel — which is exactly when a per-call bill would be spiking.
• Stays yours. The number, the recordings, the records and the workflow.

The comparison, honestly

	Ruby	Smith.ai	Built (Megabite)
Who answers	A person	AI, with live agents behind it	AI, escalating to your people
Billing basis	Per minute	Per call	Setup, monthly, then usage
Cost behaviour	Rises with call length	Rises with call volume	Largely fixed; usage scales
Start-up time	Days	Days	Weeks
Integration with your systems	Limited	Available, within the product	Built to your stack
Qualification rules	Script you provide	Configurable in-product	Yours, arbitrarily specific
Emotional or complex calls	Strongest	Escalates to an agent	Escalates to your team
Currency and base	USD, North America	USD, North America	GBP, UK and Spain

Pricing above reflects each provider's published figures at the time of writing and is given to illustrate the billing model rather than as a quotation. Check the current figures on their own sites before deciding anything.

If you are in the UK or Europe, check three things

Both Ruby and Smith.ai are North American businesses, and that is not a criticism — it is a fit question with three specific parts.

• Hours. Confirm that cover genuinely spans your working day and your out-of-hours, rather than a US business day. This is the one that catches people out, because out-of-hours is usually the reason for buying in the first place.
• Currency. Billing in dollars means your costs move with the exchange rate. On a service you intend to keep for years, that is a real variable.
• Data. Ask where call recordings and transcripts are stored and processed, and get the answer in writing. If you handle health, financial or client-confidential information, this is a compliance question and not a preference.

None of these disqualify either provider. They are simply the questions a UK buyer should ask and frequently does not.

Which one you should actually buy

Genuinely, and in the order these situations come up.

• A handful of calls a month. None of the above. A basic answering service, or let voicemail do its job.
• Low volume, and calls that genuinely need a person. Ruby. You are paying for human judgement and that is what you get.
• Moderate volume, no integration needs, live next week. Smith.ai. Capable and quick to start.
• You are losing calls out of hours, at weekends, or through every busy period. A build. This is the most common reason businesses come to us: it is the one problem more staff cannot solve, and per-call cover gets expensive precisely when volume is highest.
• The call has to end in a booking and a CRM record, not a message. A build. If your commercial next step is an appointment in a specific diary, anything short of that leaves the job unfinished.
• Your qualification is specific to your business. A build. Templates cover the common cases; the value is usually in the uncommon ones.
• Volume is high enough that per-call pricing is compounding. A build — and the arithmetic is worth doing before you renew anything.
• You want the cheapest option available. Not us. Buy message taking and spend the difference elsewhere.

The honest summary: a productised service is a good purchase while call handling is a task. Once it becomes part of how the business converts — booked appointments, qualified records, cover at the hours you are losing money — the economics and the ceiling point the same way.

Questions worth asking whoever you choose

• What exactly is a billable call? Wrong numbers, sales calls and hang-ups are often chargeable.
• What is the billing increment? Per-minute rounded up makes short calls disproportionately expensive.
• What happens outside the bundle? Overage rates matter more than the headline plan price.
• Can it book into a live calendar, or only take a message for someone to action later?
• What does it write back to your CRM, and does that need a higher tier?
• What happens when it cannot answer? The escalation path is the part that decides whether callers are served or annoyed.
• What do you keep if you leave? Recordings, transcripts, contact records and the number itself.

Common questions

What are the alternatives to Ruby Receptionists?

The realistic alternatives are other human answering services, hybrid providers such as Smith.ai, AI-only receptionist products, and having one built and integrated around your own systems. Which is right depends mostly on call volume, how complex your calls are, and whether the call has to end inside your CRM and calendar.

Is Smith.ai or Ruby cheaper?

It depends on your call pattern rather than the headline price, because they bill on different bases. Ruby charges per minute, so long calls cost more. Smith.ai charges per call, so length does not matter but volume does. Work out your average call length and monthly volume before comparing plans.

Do Ruby or Smith.ai work for UK businesses?

Both are North American businesses billing in US dollars. Rather than assume either way, ask directly about hours of cover in your timezone, the currency you will be billed in, and where call data is stored and processed.

When is building your own worth it?

When call volume is high enough that per-call or per-minute pricing is compounding against you, when out-of-hours cover is the main reason for buying, or when the call needs to end with a booking in a live diary and a record in your CRM rather than a message. Below that, a productised service is usually the better purchase.

Is an AI receptionist as good as a person?

Not for everything, and anyone claiming otherwise is selling. AI handles volume, consistency and hours better than any human rota. People handle judgement, nuance and upset callers better than any model. A sensible setup uses AI for the routine and routes the rest to a person quickly.