A small business in any major US metro deals with a multilingual customer base. Atlanta has Spanish, Vietnamese, Korean, Mandarin, French. Houston adds Punjabi, Arabic, Tagalog. New York adds everything in the world. The owner of a service business in any of those metros knows a specific cost: every phone call from a Spanish-speaking caller they cannot answer in Spanish is a lost booking. The math is brutal; the cost compounds.
The 70-language voice receptionist solves the cost. The agent detects the caller's language on the first sentence, replies in that language, in the owner's cloned voice, with the same booking and invoicing capabilities as the English flow. This post is what that looks like in real use, where the technology genuinely shines, and where the limits actually are.
What "70 languages" actually covers
The list is the major world languages plus most of the regional ones a US-metro small business would encounter. A representative slice:
European: English, Spanish, French, German, Italian, Portuguese, Dutch, Polish, Russian, Greek, Swedish, Norwegian, Danish, Finnish, Czech, Hungarian, Romanian, Ukrainian.
Asian: Mandarin, Cantonese, Japanese, Korean, Vietnamese, Thai, Indonesian, Malay, Filipino, Hindi, Bengali, Tamil, Telugu, Punjabi, Urdu, Marathi, Gujarati.
Middle Eastern + African: Arabic (multiple dialects), Hebrew, Persian, Turkish, Swahili, Amharic, Yoruba, Hausa.
Americas: Brazilian Portuguese, Mexican Spanish, Argentinian Spanish, Quebec French.
The 70 supported languages cover roughly 98% of the languages a US small business will encounter. For anything outside the supported set, the agent falls back to English with a "I'd love to help, would you mind in English?" prompt and continues from there.
How it actually works
Three layers.
- 1Language detectionOn the first sentence, the agent identifies the language with high accuracy.
- 2Voice synthesisYour cloned voice is rendered in the detected language using a multilingual model.
- 3Memory persistenceThe customer's preferred language is saved to their record. Future calls open in that language.
The voice synthesis layer is the unsung hero. A traditional translation approach would take English voice + translation engine + foreign-language TTS. The result sounds wrong: the cadence is American-English on top of Spanish words. The modern approach uses multilingual voice models that natively render the cloned voice in any of the supported languages with native-speaker prosody.
The result: a Spanish-speaking caller hears your voice, in Spanish, with Spanish cadence. They do not know it is AI; they are not even sure if it is you (until they realize you do not actually speak Spanish that well).
What this changes for businesses in multilingual metros
Three concrete shifts.
Spanish-speaking callers book at the same rate as English callers. This is the headline. Service businesses in multilingual metros typically see a meaningful conversion gap on non-English calls before the multilingual agent (often ~30% lower booking rate). With the multilingual agent, the gap closes. The numbers are real money.
You do not need to hire a Spanish-speaking receptionist. A bilingual receptionist in Atlanta costs an additional $10-15K/year over English-only. The agent removes that line item.
Customers feel respected. A Vietnamese-speaking customer in Houston who reaches a Vietnamese-speaking receptionist on a small-business line in 2026 is genuinely surprised. The bar for multilingual service in small business is low; meeting it is a competitive advantage.

Where the technology genuinely shines
Three categories where the multilingual agent matters more than baseline.
Service trades in major metros. HVAC, plumbing, electrical in Atlanta, Houston, Los Angeles, Chicago, Miami. Spanish penetration is high; many small contractors do not speak Spanish. The agent levels the playing field.
Salons and beauty in immigrant communities. Korean salons in Atlanta. Vietnamese nail salons everywhere. Arabic-speaking barbers in Detroit. The customer comfort gap of a same-language receptionist is real; the agent provides it without staffing.
Hospitality and tourism. Hotels, restaurants near tourist destinations. International callers. The agent adapts to whatever the caller speaks.
Where the limits are
Three honest gaps.
Cultural nuance beyond language. The agent speaks Mandarin. It does not necessarily understand the cultural context of a Cantonese-speaking elder requesting a specific style of haircut. The translation layer is fluent; the cultural-fluency layer is shallower. For high-stakes cultural contexts, a human receptionist still has an edge.
Code-switching mid-conversation. Most multilingual customers code-switch (English with Spanish phrases mixed in). The agent handles this acceptably but not perfectly. Heavy code-switchers sometimes feel the agent's response is too formal in one language when their input was casual mixed.
Less-common dialects. "Spanish" covers Mexican, Puerto Rican, Argentine, Castilian. The agent's default Spanish is broadly Latin American (closest to the most common US Spanish-speakers); other dialects are present but slightly weaker. This is improving over time as voice models get better.
Setup specifics
The setup wizard handles language selection in three steps.
- Detect your business's languages. From your customer-record history (if you have one), the agent suggests the languages your customers are most likely to speak.
- Confirm the languages. You confirm; the agent enables them. Some businesses pick "all 70" and let the agent auto-detect; some pick a specific 3-5.
- Test calls in each language. Call your number; speak each language; confirm the agent responds in kind. The wizard surfaces any languages where the voice clone needs more sample audio for that language specifically.
For most US-metro small businesses, the right setting is 5-10 active languages (English plus the locally-common ones). Setup is 5 minutes.
- Spanish callers get an English voicemail. ~30% don't call back.
- Owner refers Spanish callers to a competitor with a Spanish line.
- Hire a bilingual receptionist for $40-60K/year.
- Refuse to take international tourist calls.
- Spanish callers hear your voice, in Spanish, on the second ring.
- Booking rate matches English.
- No additional staff cost.
- International calls handled cleanly across 70 languages.
Frequently asked
Does my voice clone need separate samples per language?
What about the personal agent's voice across languages?
Will customers know they are talking to AI?
What if I add a new language later?
Are there any languages with worse quality?
Try it
Call your own business number and try the language test. Speak the second-most-common language your customers speak. The agent's response is the test.


