Running a two-language phone line, and the four ways to do it
There are four structures available and most businesses pick one by accident. A single bilingual assistant. Two numbers, one per language. One number with a language choice at the start. Or an English assistant with a routing rule that sends other-language callers to a person. Each fails differently, and the choice matters more than the vendor.
Two numbers is the most reliable and the least marketed, because it gives each language a clean path and no detection step to get wrong. The cost is that you have to publish both numbers everywhere and some callers will still ring the wrong one.
A language choice at the start is what most people reach for and it carries the IVR problem: a caller who has to listen to two options in two languages before speaking has already lost ten seconds and some patience. It works, and it is not free.
Routing to a person is the honest structure when your software handles one language, which is our situation. It is also the only one of the four that needs a human available, so it is a staffing decision wearing a technology decision's clothes.
How this was put together
Compared in September 2026 on how a two-language line can actually be built, by a vendor whose software covers one of them.
What we compared
- Which of the four structures each vendor supports without custom work.
- Whether a second number costs extra, which on per-agent pricing it does.
- What the caller experiences at the moment the language is decided.
- Published price and overage against a two-line setup rather than one.
What we could not test
- Which structure converts better, which would need comparable businesses running each and nobody publishes that.
- How each system behaves when the language detection is wrong, which is the failure that annoys callers most.
- Whether human services staff both languages at all hours or only in core hours.
The shortlist at a glance
Billing unit first, because it decides more of the final bill than the headline price does.
| Charges by | Entry plan | After the allowance | |
|---|---|---|---|
| GreetKeeper | per answered call | $49/mo for 200 answered calls | $0.45 a call on Starter, published |
| Dialzara | per minute | Business Lite $29/mo for 60 minutes | Published on every tier, from $0.48 a minute down to $0.35 |
| Goodcall | per unique caller | Starter $79/mo ($66 annual) for 100 unique customers | $0.50 per additional unique customer, published |
| Smith.ai | per call | Free tier at $0 for 25 calls, then $3.00 a call | Published, and the per-call rate falls as the tier rises |
| Ruby | per minute | $250/mo for 50 minutes | Not published on their pricing page |
| AnswerConnect | per minute | Does not publish pricing | Does not publish pricing |
Read from each vendor's own pricing page in September 2026.
GreetKeeper
Built for the routing structure, where one language dominates and the rest reach a named person.
Worth it for: A business whose inbound is mostly English, wanting minority-language callers sent to a colleague by rule rather than made to sit through a language menu.
Not for: The single-assistant structure, which is what many two-language businesses actually want. Our software handles English at launch and we will not claim otherwise. No customers, no reviews, no certifications, and no uptime figure.
Charges per answered call. Entry plan $49/mo for 200 answered calls. After the allowance: $0.45 a call on Starter, published.
Dialzara
Supports the single-assistant structure, which is the one that needs no routing at all.
Worth it for: A business that wants one number and one assistant covering both languages, with no detection step and nobody to staff.
Not for: Volume on the entry plan, where 60 minutes is a handful of conversations, and any structure involving a third language.
Charges per minute. Entry plan Business Lite $29/mo for 60 minutes. After the allowance: Published on every tier, from $0.48 a minute down to $0.35.
Goodcall
The two-number structure is clean here, and expensive, because a number is an agent.
Worth it for: A business running a separate line per language that values uncapped minutes on both.
Not for: Anyone using two numbers on a budget, since per-agent pricing means the second line roughly doubles the bill.
Charges per unique caller. Entry plan Starter $79/mo ($66 annual) for 100 unique customers. After the allowance: $0.50 per additional unique customer, published.
Smith.ai
The human team is what makes the route-to-a-person structure actually work.
Worth it for: A business whose software covers one language and that needs somebody reachable for the rest without hiring for it.
Not for: Assuming their team covers your specific language. That is a question to put to them directly rather than infer from a roundup.
Charges per call. Entry plan Free tier at $0 for 25 calls, then $3.00 a call. After the allowance: Published, and the per-call rate falls as the tier rises.
Ruby
People, who resolve the structure question by not having one.
Worth it for: A business that would rather pay for humans than design a routing tree, at low volume.
Not for: Volume, at $250 for 50 minutes, with no published overage rate.
Charges per minute. Entry plan $250/mo for 50 minutes. After the allowance: Not published on their pricing page.
AnswerConnect
Human cover around the clock, with the structure handled by their staffing rather than your configuration.
Worth it for: A business wanting 24/7 human answering from a long-running operator.
Not for: Published comparison of any kind, and short calls, given per-minute billing that rounds up.
AnswerConnect does not publish its rates, so the cost is whatever the sales call produces.
The structure to choose, by how your calls actually split
If the split is roughly even and both languages are constant, two numbers is the structure that fails least. Each language gets a clean path, nothing has to be detected, and the marketing cost of publishing two numbers is smaller than the conversion cost of getting detection wrong.
If one language is a clear minority, an assistant in the majority language with a routing rule for the rest is usually better than making every caller sit through a language menu. That rule needs somebody to route to, which is the part businesses forget to staff.
And if the minority language is occasional, handle it with a person and do not buy anything. Building a two-language phone system for a handful of calls a month is a large amount of configuration for a small amount of benefit.
Our own pricing, in full
Including the rate after the allowance, which is the number most of this list leaves off.
| Plan | Monthly | Calls included | After that |
|---|---|---|---|
| StarterFor a single location taking calls one line at a time. | $49 | 200 | $0.45 per call |
| ProfessionalBooking, routing and transfers included, not an add-on. | $149 | 750 | $0.35 per call |
| ScaleFor busy phones and longer opening hours. | $349 | 2,000 | $0.25 per call |
The short version
- Booking is included on every tier. Rosie keeps booking and live transfer behind its $149 plan.
- The overage rate is on the pricing page. Rosie, Ruby, Slang and AnswerConnect do not publish theirs.
- Calls are counted, not minutes, so a long useful conversation costs the same as a short one.
Questions people ask about this list
Which structure should I pick?
If both languages are constant and roughly even, two numbers. If one dominates, a single assistant in that language with a routing rule for the rest. A language menu at the start is the option that looks obvious and costs the most caller patience.
Does a second number cost extra?
With Goodcall, materially: pricing is per agent and each phone number is an agent. On most other plans a second number is a telephony cost rather than a plan cost, which is worth confirming per vendor.
What happens if language detection gets it wrong?
The caller hears a greeting they may not understand and usually hangs up, which is why the structures that avoid detection entirely fail less often than the ones relying on it.
Which structure does GreetKeeper support?
The routing one. Our software handles English at launch, so a two-language line built on us needs a person for the other language. Dialzara states support for one assistant covering both. We do not, and we will not claim a language we have not verified.
Hear one of these handle your own call
Your scenario, your greeting, a couple of minutes.