Smith.ai vs Ruby: Which Is Cheaper for a Two-Person Office?

Coding Liquids tutorial cover featuring Sagnik Bhattacharya for Smith.ai vs Ruby: Which Is Cheaper for a Two-Person Office?
Coding Liquids tutorial cover featuring Sagnik Bhattacharya for Smith.ai vs Ruby: Which Is Cheaper for a Two-Person Office?

A two-person office taking under about 65 calls a month usually pays less with Smith.ai's AI receptionist, which charges per call and is free for the first 25. Ruby's human receptionists start at $250 a month for 50 minutes. Choose Ruby when a real person on the line matters more than the saving.

The Complete Excel Course with AI Integration

From formulas and pivot tables to Power Query, DAX, VBA and AI with Copilot, ChatGPT and Claude — hands-on from the first lesson.

By Sagnik Bhattacharya and Virginia Thorn, Primetime Emmy Award winner

Buy now

The two services are priced in different units, so the headline numbers cannot be compared directly. Ruby counts minutes of conversation. Smith.ai's AI counts calls, however long they run. The only fair comparison converts both into cost per call at the call length your office actually has, and that is what the rest of this tutorial does.

What do Ruby and Smith.ai charge, according to their pricing pages?

Ruby's virtual receptionist plans on its pricing page are $250 a month for 50 minutes, $395 for 100, $720 for 200 and $1,725 for 500. The page describes live answering around the clock, call forwarding, scheduling, lead intake, payment collection and outbound call assistance as included. What it does not say is how a minute is defined, whether unused minutes roll over, what an extra minute costs, or how long the contract runs. I could not find those on the page, so treat them as questions for Ruby's sales team rather than assumptions.

Smith.ai's AI receptionist has three plans. Free costs $0 and includes 25 real calls, with extra calls at $3 each. Pro starts at $150 for 75 calls, with higher tiers of 150 and 300 calls at $270 and about $500, and overage of $2.50, $2.30 and $2.17 a call across those tiers. Enterprise starts at $500. Smith.ai also lists test and simulated calls separately from real ones, so practising your setup does not eat the allowance.

Smith.ai additionally sells human receptionists on a separate price list: Starter at $300 for 30 calls (overage $11.50 a call), Basic at $810 for 90 calls ($10.50) and Pro at $2,100 for 300 ($8.50). Billing there is per call with no minute cap. Because this tutorial is about "minutes or calls", I include these human plans in the comparison so you can see the call-based version of Ruby's service.

PlanMonthly priceAllowanceExtra use
Ruby 50$25050 minutesNot published on the page
Ruby 100$395100 minutesNot published on the page
Ruby 200$720200 minutesNot published on the page
Smith.ai AI Free$025 calls$3.00 a call
Smith.ai AI Pro (75)$15075 calls$2.50 a call
Smith.ai AI Pro (150)$270150 calls$2.30 a call
Smith.ai human Starter$30030 calls$11.50 a call
Smith.ai human Basic$81090 calls$10.50 a call

How many calls does a Ruby plan cover at three minutes each?

Divide a Ruby plan's minutes by three and you get the number of three-minute calls it covers. Fifty minutes covers about 16 calls, 100 minutes about 33 and 200 minutes about 66. A 500-minute plan covers about 166.

That gives a full-allowance cost per call of roughly $15.00 on Ruby 50 ($250 over 16.7 calls), $11.85 on Ruby 100, $10.80 on Ruby 200 and $10.35 on Ruby 500. These are the best case: they assume you use every minute. If you use half, the cost per call doubles.

Smith.ai's human plans land in the same band at full use ($10 on Starter, $9 on Basic, $7 on Pro, as the per-call fee works out when the allowance is used), but they do not care how long the call lasts. That matters for an office whose calls are long. A ten-minute call uses 3.3 times the Ruby minutes of a three-minute call and exactly the same Smith.ai call.

The AI receptionist is in a different league on price. At 75 calls a month, Free costs 25 free calls plus 50 at $3, which is $150, the same as Pro's entry price. Past 75 calls Pro is cheaper, and past about 250 calls the 300-call tier at $500 beats the 150-call tier plus overage.

If the cost-per-call sums are the slow part, try The Complete Excel Course with AI Integration on Coding Liquids.

Main example: a two-person estate agency at 40, 70 and 120 calls

Picture two negotiators in a small estate agency. One is out at viewings most of the day, the other is on the desk. They want every call answered, not just the ones that arrive while someone is free. The calls are mostly viewing requests, "is this still available?" and the odd valuation enquiry. These figures are illustrative; your call log will differ.

In a quiet month with 40 calls averaging three minutes, the office needs 120 minutes of talk time. Ruby 100 would run out, so Ruby 200 at $720 is the realistic choice, or Ruby 100 plus whatever overage Ruby charges. Smith.ai's AI on the Free plan costs 25 free calls plus 15 at $3, which is $45. Smith.ai's human Starter plan costs $300 for the first 30 calls plus 10 at $11.50, which is $415.

In a normal month with 70 calls, 210 minutes, Ruby 200 is just short, so you are at the 500-minute plan ($1,725) or paying overage. The AI Free plan costs 25 plus 45 times $3, which is $160, while Pro at 75 calls is $150 flat. Smith.ai's human Basic plan covers 90 calls at $810.

In a busy month with 120 calls, 360 minutes, Ruby needs the 500-minute plan at $1,725 unless Ruby's overage is cheap. Smith.ai AI Pro at 150 calls is $270. Smith.ai's human Basic plan is $810 plus 30 overage calls at $10.50, which is $1,125.

MonthCallsRuby (cheapest plan that fits)Smith.ai AISmith.ai human
Quiet40$720 (200 min)$45 (Free)$415 (Starter + overage)
Normal70$1,725 (500 min)$150 (Pro 75)$810 (Basic)
Busy120$1,725 (500 min)$270 (Pro 150)$1,125 (Basic + overage)

The gap is wide enough that price alone settles it for an office this size. The question becomes whether an AI voice is acceptable to this agency's callers. For a viewing request it often is. For a vendor ringing about their home, probably not, and that is a reason to route some calls to a person (see the hybrid example further down).

Why does call length matter more than call count on Ruby?

Call length matters because Ruby bills minutes, and the three-minute call assumed here is the weak point in any Ruby calculation, so test it against your own phone records before you decide. Your phone provider's call log or the voicemail app on your handset will show the length of each answered call over a month. Add up the minutes and divide by the number of calls.

Take a small recruitment agency. Calls from candidates checking on an application might run 90 seconds, while calls from a hiring manager describing a vacancy might run eight minutes. Forty candidate calls at 1.5 minutes is 60 minutes. Fifteen client calls at 8 minutes is 120 minutes. The month is 55 calls and 180 minutes, an average of 3.3 minutes, which looks like the three-minute assumption but hides a very uneven mix.

On Ruby, 180 minutes fits Ruby 200 at $720, a cost per call of about $13. On Smith.ai's AI Pro at 75 calls, the same 55 calls cost $150 or, on Free, 25 plus 30 times $3, which is $115. The recruitment agency's long calls cost Ruby customers more, while the AI plan would not notice them.

Reverse it for a marketing agency whose calls are almost all under a minute: supplier cold calls, "wrong number", sales pitches. Sixty of those a month at 45 seconds is only 45 minutes, which fits Ruby 50 at $250. That is $4.17 a call, not $15, because the minute allowance is barely used by short calls. Even then, Smith.ai's AI at $105 for 60 calls (25 free plus 35 at $3) is cheaper, though a lot of those short calls are the ones the AI's spam filtering would remove from the count.

Do spam and wrong-number calls count towards Smith.ai's bill?

Smith.ai says known spammers are filtered before the AI answers, and you can strike up to 10 per cent of calls from your bill each cycle for selected spam. Spam and wrong numbers are the quiet cost in a per-call plan. If a quarter of your inbound calls are robocalls, your effective bill is lower than the raw count suggests.

On Smith.ai's human plans the pricing page says wrong numbers and spam do not count towards the quota provided the forwarded line passes caller ID. That detail is worth testing: if your phone system strips caller ID when forwarding, spam may count. Make five test calls from a mobile that hides its number and one that shows it, and check how each appears in the dashboard.

Ruby's pricing page does not describe how spam is handled at all. Ask whether a robocall that connects for six seconds is billed as one minute, and whether partial minutes round up.

What should you ask Ruby before you pay?

Ask Ruby what counts as a minute, what an extra minute costs, and what the contract terms are before you pay.

These are questions, not criticisms. The answers decide which side of the table your office falls on.

  • What counts as a minute? Ask if hold time, transfer time and ring time before a handover are billed, and whether minutes are rounded up to the next full minute.
  • What is the price of a minute over the plan? The pricing page lists four plans and no overage rate. Ask for it in writing and calculate your busy-month bill using it.
  • What are the terms? Contract length, how to move between plans and what happens to unused minutes. Smith.ai's page states month-to-month with no long contract for its human plans and a 30-day money-back guarantee for new clients on the plan and add-ons up to $1,000, excluding overage. Ruby's page, as I read it, was silent.

Can a dental practice use AI for routine calls and humans for sensitive ones?

A dental practice can use a Smith.ai AI receptionist as the default answerer and route sensitive calls, such as a patient in pain, to a person.

A two-dentist practice has two kinds of call. Appointment requests and opening-hours questions are routine. Calls from a patient in pain, or about a result, are not. Putting every call through a human service is expensive and putting every call through AI is risky.

The practical blend is a Smith.ai AI receptionist as the default answerer, with a rule that any call the AI cannot resolve, or any caller who asks for a person, is taken as a message with a callback by the practice. Smith.ai's pages describe lead qualification, call routing and scheduling on the free plan. For a human voice on the sensitive calls, Smith.ai also sells hybrid and human-first plans, which would keep one vendor and one number.

Before relying on it, make 20 test calls: ten routine and ten "awkward" ones (an angry caller, a mumbled name, someone asking if the dentist is available right now). Smith.ai separates test calls from real ones, so the practice's allowance is not used up. Listen to the recordings, note where the AI went wrong and either fix the instructions or move that call type to a person.

Which plan a marketing agency's mix would pick

Take a two-person marketing agency that gets about 50 calls a month: 30 are prospects, 10 are existing clients and 10 are suppliers or sales pitches. Prospects and clients average four minutes, the rest 40 seconds. That is 160 minutes plus about seven, so roughly 167 minutes of talk time.

On Ruby, 167 minutes means the 200-minute plan at $720, or $14.40 a call. On Smith.ai's AI the same 50 calls cost 25 free plus 25 at $3, which is $100, or $2 a call. The $620 monthly difference is real money for two people. The question is whether prospects mind. A prospect leaving an enquiry with an AI that books a call back is usually fine, and a client with an urgent issue can be routed to a mobile number. If the agency sells on warmth of service, it might decide that the 30 prospect calls are the ones worth putting a human on, which is a case for Smith.ai's human plans at 30 calls ($300), with the other 20 handled by the owners.

How should a two-person office decide between Ruby and Smith.ai?

A two-person office should count its answered calls and minutes for a month, then price Smith.ai's AI Free plan first if calls are under about 65. If more than about one call in five needs a human, price Ruby 100 or 200.

Work from your own numbers, not from the vendor's headline price.

  1. Pull a month of call history and count the answered calls and total minutes.
  2. If calls are under about 65, price the Smith.ai AI Free plan first. At 65 calls it costs $120 (40 calls over the free 25, at $3), which is $30 less than Pro's $150 entry price.
  3. Between about 75 and 250 calls, price Smith.ai AI Pro.
  4. List the call types that need a human: complaints, sensitive health, large-value enquiries. If they are more than about one call in five, price Ruby 100 or 200 and ask for overage terms, or look at Smith.ai's hybrid options.
  5. Run 20 test calls on the cheapest candidate before committing.

The scenario above shows the weak spot of the two-person office: calls arrive unpredictably, and the plan has to cover the busiest month. A flat plan sized for a quiet month punishes you with overage in a busy one. Smith.ai's per-call tiers scale more smoothly than Ruby's four minute steps, and the free plan means a quiet month costs almost nothing. For more on the tier maths, Smith.ai pricing at 50, 150 and 300 calls goes tier by tier, the three-way price comparison with Rosie adds another per-call option, and the 60-call comparison of Smith.ai's AI and human plans zooms in on a typical month.

What does a wrong choice look like after a month?

A wrong choice shows up within a month as a Ruby bill above the plan price, or as callback requests from callers the AI could not help. The first failure is buying Ruby 50 at $250 because it looks cheap, then discovering that 90 minutes of calls came in. If Ruby's overage rate is high, the real bill can exceed the next plan up. The tell is the bill arriving above the plan price in month one.

The second is choosing the AI plan on price and never listening to the calls. The tell is a pile of callback requests that begin "I rang earlier and nobody could help." Read ten recordings a week for the first month and keep a tally of calls the AI handled well, handled badly and should have passed to a person. If "badly" is more than one in ten, move those call types elsewhere.

Check the Smith.ai dashboard count against your own phone log in the first fortnight. They should match within the spam you expect to be filtered. If they do not, forwarding is probably sending duplicates, for example both a mobile and a desk phone forwarding into the same number.

How do you check the numbers before signing with either vendor?

Check the numbers before signing by running through the list below with both vendors' pricing pages open. It takes about 20 minutes.

  • Open ruby.com/pricing and smith.ai/pricing and confirm the four Ruby prices and Smith.ai's AI tiers still match what is quoted here. Plans change, and so do promotions.
  • Check your own call log for the last three months, not one. A single month can mislead.
  • Count how many of the calls were repeat callers or spam. Those lower the AI bill but may raise a minute-based bill.
  • Decide the fallback: what happens if the receptionist cannot answer a question. Both vendors can take messages and route calls, but the human route differs from the AI route.
  • Write down the cancellation terms in your own words and keep the page's date.

If you would rather work this out against your actual call log, that is the kind of job my AI implementation consultation is for. Either way, run the sums yourself first: the whole comparison turns on two numbers, calls a month and minutes a call.

Smith.ai and Ruby: questions a two-person office asks

Does Ruby publish what happens when I go over my minutes?

Not on Ruby's pricing page at the time of writing. It lists the four minute plans and their monthly prices but gives no overage rate and no definition of a minute. Ask Ruby's sales team in writing whether hold time and transfers count and what an extra minute costs before you sign.

Is Smith.ai's AI receptionist a person answering?

No. The AI receptionist plans (Free, Pro and Enterprise) are answered by AI, not a person. Smith.ai also sells separate human-receptionist plans, from $300 a month for 30 calls, which are priced differently and covered in this tutorial's comparison table.

Do spam calls count against my Smith.ai AI allowance?

Smith.ai says calls from known spammers are filtered out before the AI answers and do not count, and you can remove up to 10 per cent of calls from your bill each cycle for selected spam calls. Every other answered call counts by default.

Which is better if clients expect a human voice?

Ruby, or Smith.ai's human plans, if clients expect a human voice. An AI receptionist can book, qualify and take messages, but some clients, such as someone calling about a sensitive matter, still want a person. Decide by who calls you and why, not only by price.

Can I mix them, with AI for routine calls and humans for the rest?

Yes, within one vendor. Smith.ai describes AI-first, hybrid and human-first options on its pricing pages, so a blend exists inside one vendor. Mixing Ruby with another AI service means two numbers or two forwarding rules, which a two-person office has to maintain.

How much does Smith.ai's AI receptionist cost per call?

Smith.ai's AI Free plan costs $0 and includes 25 real calls, with extra calls at $3 each. Pro starts at $150 for 75 calls, with higher tiers of 150 and 300 calls at $270 and $500, and overage of $2.50, $2.30 and $2.17 a call across those tiers. Enterprise starts at $500.

How much does Ruby cost per call at three minutes a call?

At three minutes a call, a Ruby plan costs roughly $15.00 a call on Ruby 50, $11.85 on Ruby 100, $10.80 on Ruby 200 and $10.35 on Ruby 500, if you use every minute. If you use half, the cost per call doubles. Divide the plan's minutes by three to see how many calls it covers.

Further reads

Sources: Ruby's pricing page (ruby.com/pricing), Smith.ai's AI receptionist and receptionist pricing pages (smith.ai/pricing), checked 17 Nov 2026. Prices are USD list prices and can change.

Want a second pair of eyes on your setup?

Bring the process you're trying to fix. On a 1:1 call we'll work out which AI tools should handle it, which parts stay manual, and how to roll it out without disrupting the team.

Book a 1:1 call with me